Integrated Photonics for the AI Revolution - An Ecosystem Perspective

Full Transcript

https://www.youtube.com/watch?v=ke4Dqcjl3lo

[00:00] Hi everyone, I'm Melissa Callas of the.
[00:03] Hi everyone, I'm Melissa Callas of the Advanced Photonics Coalition.
[00:06] Welcome to our webcast, an ecosystem perspective of integrated photonics for the AI revolution.
[00:09] Integrated photonics for the AI revolution.
[00:13] This webcast is an outreach to everyone watching this video.
[00:16] We must collaborate and create standards in order to move our industry forward and to scale pck technology to volumes.
[00:19] Collaborate and create standards in order to move our industry forward and to scale pck technology to volumes.
[00:22] The only way this is going to happen is if all of us work together.
[00:24] And the Advanced Photonics Coalition provides the neutral ground to work with.
[00:28] This webcast will also be a live panel at OFC in San Diego.
[00:30] So we encourage you to come out and see this live in person in March.
[00:32] I'd like to thank all the participants of this video, including Jose Poo of Optica for hosting it and Nancy Lima of the DuPont Silicon Valley Technical Center for producing it.
[00:35] Jose Poo of Optica for hosting it and Nancy Lima of the DuPont Silicon Valley Technical Center for producing it.
[00:36] The
[01:02] Technical Center for producing it the slides are available now at advanced-c org.
[01:08] If you'd like to take a closer look at those and we thank you so much for joining us.
[01:19] My name is Koso I'm the chief technology officer of Optica the global photonic and Optics industry and academic organization and had the honor and pleasure to be here today with you for a very important meeting.
[01:34] Over the next two hours we are going to have ah witnessing of how the photonic industry is addressing the AI Revolution.
[01:44] The this meeting is organized by the advanced photonics Coalition and in a couple of minutes you're going to hear about what this is and how it can benefit your business.
[01:54] Being part of the advanced photonic Coalition is something that Optica endorses for developing your business in the field of of Optics and photonics for Datacom but on the next
[02:02] photonics for Datacom but on the next two hours listen to this we wanted to.
[02:04] two hours listen to this we wanted to have the entire value chain represented.
[02:07] have the entire value chain represented for the out of for the AI Revolution for.
[02:11] for the out of for the AI Revolution for example for the fundry we're going to.
[02:13] example for the fundry we're going to have Global fund represented by Vias.
[02:16] have Global fund represented by Vias Gupta for the osat we're going to have.
[02:18] Gupta for the osat we're going to have silitronics with Das Bora as a contract.
[02:22] silitronics with Das Bora as a contract manufacturer the company that everyone.
[02:25] manufacturer the company that everyone everyone is talking about we are going.
[02:27] everyone is talking about we are going to have Jill represented by no fear and.
[02:30] to have Jill represented by no fear and also we wanted to understand the views.
[02:32] also we wanted to understand the views from the equipment manufacturer.
[02:35] from the equipment manufacturer perspective and for that we have no.
[02:37] perspective and for that we have no other that the CEO of ficon tech Toren.
[02:40] other that the CEO of ficon tech Toren ven camp and since one of the challenges.
[02:43] ven camp and since one of the challenges perhaps the main challenge is testing.
[02:46] perhaps the main challenge is testing qu5 photonics will be there represented.
[02:48] qu5 photonics will be there represented by my very good friend case propstra and.
[02:52] by my very good friend case propstra and in every value chain you need to have.
[02:54] in every value chain you need to have the customer and for that we are honored.
[02:57] the customer and for that we are honored with the pressure of manage mecta from.
[02:59] with the pressure of manage mecta from broadco this event is going to take.
[03:01] broadco this event is going to take exactly two hours of your time and every.
[03:03] exactly two hours of your time and every minute will count.
[03:06] I would like to remind you that you have any questions that you should have questions please use the Q&A button and always in your question ask answer the person that you want the question to be addressed to.
[03:18] Very important, please tell us who in the room should answer such question or we're going to have Place plenty of questions.
[03:25] And after this event we're going to have a panel at ofc and you're going to hear more about that later on but now I would like to give the floor to the chairman of the advanced photonic Coalition my personal friend and a person who has meant the world to First the Kobo Alliance now APC and who has a passion to bring companies together.
[03:47] Jeff the floor and the attention of everyone is yours.
[03:51] Tell us what I BC.
[03:58] Hello, yes, let me uh share my presentation.
[04:05] we can see it just go presentation mode.
[04:08] we can see it just go presentation mode.
[04:13] yeah it takes time all right so I'm Jeffrey Macky vice president and treasur.
[04:16] Jeffrey Macky vice president and treasur of advanced phonics Coalition uh we are.
[04:19] of advanced phonics Coalition uh we are previously known as Kobo but today we are APC.
[04:22] previously known as Kobo but today we are APC uh the purpose of the.
[04:25] are APC uh the purpose of the corporation is to improve business conditions for and.
[04:27] corporation is to improve business conditions for and uh sorry my screen is not behaving.
[04:30] uh sorry my screen is not behaving properly yeah so uh business conditions for and promote the common business.
[04:36] properly yeah so uh business conditions for and promote the common business interest of persons developing manufacturing integrating and using.
[04:39] for and promote the common business interest of persons developing manufacturing integrating and using optical interconnects and Optical.
[04:40] interest of persons developing manufacturing integrating and using optical interconnects and Optical subassemblies in network equipment for.
[04:42] manufacturing integrating and using optical interconnects and Optical subassemblies in network equipment for data center high performance Computing.
[04:44] optical interconnects and Optical subassemblies in network equipment for data center high performance Computing Quantum Computing artificial intelligence and machine learning.
[04:46] subassemblies in network equipment for data center high performance Computing Quantum Computing artificial intelligence and machine learning applications.
[04:48] data center high performance Computing Quantum Computing artificial intelligence and machine learning applications so that's our purpose you can find that in our corporate uh document.
[04:50] Quantum Computing artificial intelligence and machine learning applications so that's our purpose you can find that in our corporate uh document.
[04:52] artificial intelligence and machine learning applications so that's our purpose you can find that in our corporate uh document.
[04:53] intelligence and machine learning applications so that's our purpose you can find that in our corporate uh document.
[04:56] applications so that's our purpose you can find that in our corporate uh document but really our mission then is for.
[04:59] can find that in our corporate uh document but really our mission then is for.
[05:00] document but really our mission then is for working with Optical Fabs so that they.
[05:02] but really our mission then is for working with Optical Fabs so that they can be leading and creating the world's.
[05:04] working with Optical Fabs so that they can be leading and creating the world's.
[05:06] can be leading and creating the world's best technology in terms of reliability.
[05:08] best technology in terms of reliability validation and trust trust being the.
[05:11] validation and trust trust being the collaboration to achieve those.
[05:13] collaboration to achieve those objectives so this allows for the.
[05:15] objectives so this allows for the creation of building blocks for silicon.
[05:17] creation of building blocks for silicon photonics I that are put together in an.
[05:19] photonics I that are put together in an assured way using Foundry process.
[05:21] assured way using Foundry process development kit to accelerate production.
[05:24] development kit to accelerate production okay that all sounds good but we need.
[05:27] okay that all sounds good but we need more than that what we see in the.
[05:30] more than that what we see in the industry is an ecosystem that achieves.
[05:33] industry is an ecosystem that achieves that and we have different levels in the.
[05:35] that and we have different levels in the ecosystem we have the hyperscalers and.
[05:38] ecosystem we have the hyperscalers and service providers that uh are the end.
[05:41] service providers that uh are the end customers uh and some of these are also.
[05:44] customers uh and some of these are also functioning themselves as system.
[05:45] functioning themselves as system companies putting out.
[05:48] companies putting out requirements for technology and then we.
[05:50] requirements for technology and then we have the traditional system companies.
[05:53] have the traditional system companies and as we drop down through this.
[05:54] and as we drop down through this ecosystem we now have the the photonic.
[05:57] ecosystem we now have the the photonic Founders playing a very important role.
[05:59] Founders playing a very important role for our silic photonics uh components.
[06:03] for our silic photonics uh components and module builds and so on and so in.
[06:07] and module builds and so on and so in this ecosystem we have modules uh the.
[06:10] this ecosystem we have modules uh the osat and uh active and passive.
[06:14] osat and uh active and passive components so this entire ecosystem at.
[06:17] components so this entire ecosystem at every level of it can benefit from.
[06:20] every level of it can benefit from standards standards help to drive access.
[06:23] standards standards help to drive access to.
[06:23] to technology and uh also multi-sourcing.
[06:28] technology and uh also multi-sourcing capability so that we can as an industry.
[06:31] capability so that we can as an industry drive volume the whole idea here is to.
[06:34] drive volume the whole idea here is to uh break away from silos we'd say silic.
[06:37] uh break away from silos we'd say silic phonics then it'd be like well which one.
[06:40] phonics then it'd be like well which one and that means then.
[06:42] and that means then the access to Silicon bonics uh was.
[06:46] the access to Silicon bonics uh was limited in volume because we only have.
[06:48] limited in volume because we only have the access uh of that one particular.
[06:51] the access uh of that one particular silo so now if we can work at each level.
[06:54] silo so now if we can work at each level of this ecosystem to use standards to uh.
[06:58] of this ecosystem to use standards to uh break out of these silos.
[07:00] break out of these silos we can now drive a larger volume and.
[07:03] we can now drive a larger volume and cost reduction for the industry and the.
[07:06] cost reduction for the industry and the commonality then hopefully will bring us.
[07:09] commonality then hopefully will bring us to a very vibrant industry different.
[07:11] to a very vibrant industry different than we've been seeing so our whole.
[07:13] than we've been seeing so our whole premise now at APC is to be helping.
[07:16] premise now at APC is to be helping every level of the ecosystem to.
[07:18] every level of the ecosystem to recognize where the opportunity of.
[07:20] recognize where the opportunity of Standards uh has been but not yet um.
[07:24] Standards uh has been but not yet um exercised and and indeed exercise that.
[07:28] exercised and and indeed exercise that opportunity so we have a number of.
[07:31] opportunity so we have a number of companies that are a member of of APC.
[07:34] companies that are a member of of APC and we're growing in the in the number.
[07:36] and we're growing in the in the number of members and so looking forward then.
[07:39] of members and so looking forward then to uh uh hearing from more companies.
[07:43] to uh uh hearing from more companies that wish to join and today are many of.
[07:46] that wish to join and today are many of our presenters are already members of.
[07:48] our presenters are already members of APC so thank you for your attention and.
[07:51] APC so thank you for your attention and I hope I have uh made it clear we're.
[07:54] I hope I have uh made it clear we're we're here to help thank you thank thank.
[07:57] we're here to help thank you thank thank you very much Jeff as I said before for.
[07:58] you very much Jeff as I said before for me it is very important that you.
[08:00] me it is very important that you understand the value of APC many of you.
[08:03] understand the value of APC many of you are corporate members of optica APC is.
[08:06] are corporate members of optica APC is here to make an iteration on connecting.
[08:09] here to make an iteration on connecting the supply chain address challenges go.
[08:11] the supply chain address challenges go from the de facto standards all the way.
[08:13] from the de facto standards all the way to understand the challenges one part of.
[08:15] to understand the challenges one part of the supply chain to the needes of the.
[08:17] the supply chain to the needes of the next one and that is why we are meeting.
[08:19] next one and that is why we are meeting together to have an example of a supply.
[08:22] together to have an example of a supply chain and understand how a company will.
[08:25] chain and understand how a company will have potential suppliers and customers.
[08:27] have potential suppliers and customers and how they can work together to solve.
[08:29] and how they can work together to solve the overall AI Challenge and to start.
[08:32] the overall AI Challenge and to start with let's start with a company that.
[08:35] with let's start with a company that really has meant the success of silicon.
[08:38] really has meant the success of silicon photonics when Global fund said to.
[08:40] photonics when Global fund said to everyone we offer silicon photonic.
[08:42] everyone we offer silicon photonic production was when companies realized.
[08:44] production was when companies realized that the volume production demanded by.
[08:46] that the volume production demanded by the applications was there when they.
[08:48] the applications was there when they decide to go into Quantum technology Z.
[08:50] decide to go into Quantum technology Z companies realized that Quantum.
[08:52] companies realized that Quantum technology was something serious today.
[08:54] technology was something serious today we are here to understand how silicon.
[08:56] we are here to understand how silicon photonics can address the AI challenge.
[08:58] photonics can address the AI challenge known order that Global fundra we be.
[09:00] known order that Global fundra we be represented by vas kta vas thank you.
[09:03] represented by vas kta vas thank you very much for being with us today the.
[09:05] very much for being with us today the floor and the attention of everyone goes.
[09:07] floor and the attention of everyone goes to Global fundies the floor is yours.
[09:09] to Global fundies the floor is yours thank you very much Jeff uh let me share.
[09:12] thank you very much Jeff uh let me share my let me share my presentation uh hopefully everybody can see it uh I'm going to put it in okay uh so hi my name is uh vikas Gupta I'm the senior director of product management focused on Silicon photonics at Global foundaries and today I'm going to use my 10 minutes uh to give give a perspective on The Foundry and an ecosystem view of silicon photonics so let me just start by making a statement that ative AI model size the compute systems that are required to actually run these models are also becoming very complex and and one of the key sort of items that is holding some of these system back is the interconnect and and the interconnect basically the the requirements are you know high energy efficiency low loss and high speed and silicon photonics as a
[10:14] high speed and silicon photonics as a technology uh can meet these requirements.
[10:17] So what I'm going to do is cover two topics uh just give you all a brief overview of GF photonics.
[10:22] That is the brand name of the photonics technology, silicon photonics technology that Global Foundries offers.
[10:26] And then I put it in p uh and and what GF is doing as a company.
[10:34] But then also what the industry needs to step up with respect to uh the ecosystem.
[10:39] So let me just start off by overview of the photonics technology that Global Foundries offers.
[10:48] So you can think about GF photonics as a bit of a system on a chip.
[10:50] And what we've done is we've integrated electronics and photonics on a 300 mm wafer sSOI wafer.
[11:00] And so you can see in this diagram on the left uh the electronic components represented by the nfet and the pfet.
[11:08] But then we also have precharacterized devices such as active devices such as modulators and detectors on the right.
[11:17] modulators and detectors on the right there is a fairly large suite of uh.
[11:20] there is a fairly large suite of uh silicon wave guides and silicon nitrite.
[11:23] silicon wave guides and silicon nitrite wave guides we've also introduced the.
[11:26] wave guides we've also introduced the whole concept of undercut for our.
[11:28] whole concept of undercut for our heaters to make the heaters uh more.
[11:31] heaters to make the heaters uh more efficient and in addition to that uh we.
[11:34] efficient and in addition to that uh we have Technologies for light in light out.
[11:36] have Technologies for light in light out our plan of reca technology is a vgroup.
[11:38] our plan of reca technology is a vgroup based passive fiber attach uh we're.
[11:42] based passive fiber attach uh we're going to upgrade uh this vgroup based.
[11:44] going to upgrade uh this vgroup based fiber attach to include a silicon.
[11:46] fiber attach to include a silicon nitride uh spot size converter uh for.
[11:49] nitride uh spot size converter uh for high high uh power requirements and.
[11:53] high high uh power requirements and there is work in terms of R&D in terms.
[11:55] there is work in terms of R&D in terms of attaching an on die laser uh so all.
[11:59] of attaching an on die laser uh so all of this provides sort of all of the.
[12:01] of this provides sort of all of the building blocks that you know a company.
[12:03] building blocks that you know a company would need to be able to produce uh.
[12:07] would need to be able to produce uh high-speed Optical links now in addition.
[12:10] high-speed Optical links now in addition to these components device components.
[12:12] to these components device components we've also paid attention to packaging a.
[12:16] we've also paid attention to packaging a and so towards that we have included.
[12:18] and so towards that we have included things like copper pillar we're working.
[12:20] things like copper pillar we're working on tsvs and the ability to do on wafer testing.
[12:23] on tsvs and the ability to do on wafer testing and I will go into some of those details.
[12:25] testing and I will go into some of those details uh a little more into the presentation now.
[12:27] details uh a little more into the presentation now one of the things that I wanted to point out is that we have taken a platform approach to silicon photonics.
[12:29] presentation now one of the things that I wanted to point out is that we have taken a platform approach to silicon photonics.
[12:31] I wanted to point out is that we have taken a platform approach to silicon photonics.
[12:34] taken a platform approach to silicon photonics and what that means is that we have these pre-characterized devices as part of our pdk with the intention of doing two things.
[12:36] photonics and what that means is that we have these pre-characterized devices as part of our pdk with the intention of doing two things.
[12:38] have these pre-characterized devices as part of our pdk with the intention of doing two things one is you know getting things first time right and the second thing is ease and speed speed up the introduction of products uh from that.
[12:41] doing two things one is you know getting things first time right and the second thing is ease and speed speed up the introduction of products uh from that.
[12:43] things first time right and the second thing is ease and speed speed up the introduction of products uh from that our customers may be designing to now.
[12:45] thing is ease and speed speed up the introduction of products uh from that our customers may be designing to now.
[12:49] introduction of products uh from that our customers may be designing to now the flexibility of this technology comes to the Forefront when we start talking about different architectures that can be implemented on this technology with almost no modification.
[12:51] our customers may be designing to now the flexibility of this technology comes to the Forefront when we start talking about different architectures that can be implemented on this technology with almost no modification.
[12:53] the flexibility of this technology comes to the Forefront when we start talking about different architectures that can be implemented on this technology with almost no modification.
[12:56] the flexibility of this technology comes to the Forefront when we start talking about different architectures that can be implemented on this technology with almost no modification.
[12:58] to the Forefront when we start talking about different architectures that can be implemented on this technology with almost no modification so typically you know high-speed interconnects can be divided into sort of two categories.
[13:00] about different architectures that can be implemented on this technology with almost no modification so typically you know high-speed interconnects can be divided into sort of two categories.
[13:04] be implemented on this technology with almost no modification so typically you know high-speed interconnects can be divided into sort of two categories the fast and narrow which requires electrical scaling and the slow and wide which is sort of optical scaling the
[13:07] almost no modification so typically you know high-speed interconnects can be divided into sort of two categories the fast and narrow which requires electrical scaling and the slow and wide which is sort of optical scaling the
[13:09] know high-speed interconnects can be divided into sort of two categories the fast and narrow which requires electrical scaling and the slow and wide which is sort of optical scaling the
[13:12] divided into sort of two categories the fast and narrow which requires electrical scaling and the slow and wide which is sort of optical scaling the
[13:14] fast and narrow which requires electrical scaling and the slow and wide which is sort of optical scaling the
[13:16] electrical scaling and the slow and wide which is sort of optical scaling the
[13:18] which is sort of optical scaling the electrical scaling basically requires
[13:21] electrical scaling basically requires you to have a high-speed electrical
[13:23] you to have a high-speed electrical signal uh that electrical signal is
[13:25] signal uh that electrical signal is imprinted on a laser output and and your
[13:28] imprinted on a laser output and and your your transmitting uh that same
[13:30] your transmitting uh that same electrical the same speed of that
[13:32] electrical the same speed of that electrical signal onto the fiber in
[13:35] electrical signal onto the fiber in terms of the slow and wide you can think
[13:37] terms of the slow and wide you can think of that as dwdm or dense wave Division
[13:41] of that as dwdm or dense wave Division multiplexing and you're basically taking
[13:43] multiplexing and you're basically taking a low low speed electrical signal sort
[13:46] a low low speed electrical signal sort of parallel uh electrical signal mxing
[13:49] of parallel uh electrical signal mxing all of those signals onto a fiber using
[13:53] all of those signals onto a fiber using uh different wavelengths of the of the
[13:56] uh different wavelengths of the of the laser so so the beauty of this platform
[13:59] laser so so the beauty of this platform approach is the fact that you can you
[14:02] approach is the fact that you can you sort of Mel these two architectures and
[14:04] sort of Mel these two architectures and produce these two architectures with
[14:06] produce these two architectures with very little in terms of customization of
[14:09] very little in terms of customization of the technology and all of these
[14:11] the technology and all of these components uh like the vroo you know the
[14:14] components uh like the vroo you know the ability whether you want a modulator
[14:17] ability whether you want a modulator which is a ring modulator or a mark
[14:18] which is a ring modulator or a mark sender modulator and all of the sender modulator and all of the components that you would be required to components that you would be required to put together these sort of circuits are put together these sort of circuits are part of the pdk and part of the offering part of the pdk and part of the offering that GF uh offers so with that uh I'm that GF uh offers so with that uh I'm I'm going to sort of I just wanted to I'm going to sort of I just wanted to give you all a flavor of the technology give you all a flavor of the technology that GF offers but I want to focus most that GF offers but I want to focus most of my presentation about the ecosystem of my presentation about the ecosystem that that GF is working towards but then that that GF is working towards but then also demands that we may have from the also demands that we may have from the industry so towards that let me just industry so towards that let me just start off with the definition of start off with the definition of ecosystem it is a complicated system it ecosystem it is a complicated system it requires you know different companies to requires you know different companies to work together each company may be work together each company may be specialized in their their own uh sort specialized in their their own uh sort of field of view but then it also there of field of view but then it also there is there's a high level of is there's a high level of interdependency between uh between these interdependency between uh between these companies and so what I've tried to do companies and so what I've tried to do in this chart is just show you sort of in this chart is just show you sort of the the larger context of all of the the the larger context of all of the players that potentially may be involved players that potentially may be involved some of them will be talking.
[15:27] Today, today I think we have lost connection with the.
[15:33] I think we have lost connection with the global fundies or leading fundry facility.
[15:36] Let's wait a couple of minutes.
[15:38] But in the meantime, let me remind all our audience for any questions that you have, please use the Q&A button.
[15:42] And there, write the person you want to direct the question to.
[15:45] At the end of the meeting, exactly at at 11:22 Eastern, we'll have a Q&A session.
[15:49] Which we'll address every single question that you ask, and that's always the most exciting part of the event.
[15:53] Pasas, are you with us?
[15:55] I think we are, we are in this moment, which this is the moment in which we excel on making sure that people understand this is a live event.
[15:56] And in life event, things happen, and this is why we need a more reliable internet connection for the world.
[15:58] And that's what we're trying to solve here.
[16:01] In the meantime, let me continue with the program, and we will come back to Vas in a second from Global Fies.
[16:03] But now we started with The Foundry, and we are going to continue with the oat why.
[16:30] going to continue with the oat why because the osad is the one that gets
[16:31] because the osad is the one that gets the chips from The Foundry will get the
[16:33] the chips from The Foundry will get the equipment from F cont and other
[16:35] equipment from F cont and other equipment suppliers and will provide
[16:38] equipment suppliers and will provide package devices it's the next step in
[16:39] package devices it's the next step in the supply chain for the osat we went to
[16:42] the supply chain for the osat we went to a company that is interacting in many
[16:45] a company that is interacting in many different markets data con lier we go to
[16:48] different markets data con lier we go to citronics thank you very much Das Bora
[16:51] citronics thank you very much Das Bora and Cassandra for being with us today
[16:53] and Cassandra for being with us today tell us how you take C from Global
[16:56] tell us how you take C from Global fundra and you the next step in the
[16:58] fundra and you the next step in the supply CH the floor and the attention of
[17:00] supply CH the floor and the attention of everyone is yours perfect perfect thank
[17:03] everyone is yours perfect perfect thank you Jose and Optica team for putting
[17:07] you Jose and Optica team for putting together such a great webinar and the
[17:10] together such a great webinar and the advanced photonics Coalition they
[17:12] advanced photonics Coalition they organizing the event today Advanced
[17:14] organizing the event today Advanced photonics Coalition thank you for
[17:16] photonics Coalition thank you for inviting stonic uh we are uh proud of
[17:20] inviting stonic uh we are uh proud of being a part of the Great presenters
[17:23] being a part of the Great presenters here um um Global found already talk
[17:27] here um um Global found already talk about the pick we are talking about uh
[17:30] about the pick we are talking about uh osets and what is uh needed from uh oset.
[17:35] osets and what is uh needed from uh oset point of point of view so we are going.
[17:37] point of point of view so we are going to talk about the challenges uh in um.
[17:41] to talk about the challenges uh in um integrated phon assembly I'm going to.
[17:43] integrated phon assembly I'm going to give the initial introduction and then.
[17:45] give the initial introduction and then cassendra is going to talk about the uh.
[17:48] cassendra is going to talk about the uh inuring uh the outline of our uh project.
[17:52] inuring uh the outline of our uh project is uh that we are going to talk about.
[17:55] is uh that we are going to talk about quickly what is the osed model uh and.
[17:58] quickly what is the osed model uh and then what ctronics has done so far in.
[18:00] then what ctronics has done so far in terms of the uh Market we'll quickly.
[18:03] terms of the uh Market we'll quickly touch on the Outlook what are the.
[18:05] touch on the Outlook what are the challenges when we are talking about the.
[18:08] challenges when we are talking about the photonic assembly and how Electronics.
[18:10] photonic assembly and how Electronics engineering team has come up with couple.
[18:12] of solutions which will enable uh.
[18:15] photonic uh you know uh pick assembly.
[18:19] photonic uh you know uh pick assembly faster which a much higher yield and.
[18:21] faster which a much higher yield and then we'll talk about uh the summary.
[18:24] then we'll talk about uh the summary quickly uh normally Outsourcing means uh.
[18:27] quickly uh normally Outsourcing means uh oet means Outsourcing with a third party.
[18:30] oet means Outsourcing with a third party and normally these are up in Far East.
[18:33] and normally these are up in Far East the role of oset is you know optimize.
[18:36] the role of oset is you know optimize the process in terms of yield cycle time.
[18:40] the process in terms of yield cycle time uh and they take advantage of economic.
[18:42] uh and they take advantage of economic scale to bring down the cost with the.
[18:44] scale to bring down the cost with the normal oset uh you know in the in ic.
[18:47] normal oset uh you know in the in ic they do the IC packaging whether it's.
[18:49] they do the IC packaging whether it's flip chip wire burning then the test uh.
[18:52] flip chip wire burning then the test uh and then obviously inspection during the.
[18:54] and then obviously inspection during the assembly and before the uh final.
[18:57] assembly and before the uh final shipment on the Optical uh is uh you.
[19:00] shipment on the Optical uh is uh you know still a growing market so we have.
[19:03] know still a growing market so we have pick we got laser we got wave guides so.
[19:06] pick we got laser we got wave guides so all these need to be integrated together.
[19:08] all these need to be integrated together in terms of the package and then we need.
[19:10] in terms of the package and then we need to do pre and post uh testing while.
[19:13] to do pre and post uh testing while you're doing the active alignment uh and.
[19:15] you're doing the active alignment uh and the inspection is very very key uh.
[19:18] the inspection is very very key uh because you are talking about the lasers.
[19:20] because you are talking about the lasers you're talking about the wave guide.
[19:22] you're talking about the wave guide where and then the fivers where you have.
[19:24] where and then the fivers where you have to do multiple inspection at incoming.
[19:26] to do multiple inspection at incoming during the process and outgoing uh any.
[19:29] during the process and outgoing uh any little scratch uh in the optical path.
[19:31] Little scratch uh in the optical path obviously will affect the power and the performance.
[19:33] So in a in a new o set where you're talking about the optical IC and the photonix IC uh that's where uh we are going to focus today uh in terms of what is the treny uh service.
[19:44] So treny means uh design fabrication assembly and test all being under one roof for oset.
[19:55] And normally oset will take over when the the process is mature and is ready for high volume for silicon photonic.
[20:03] As we all know Market is developing is growing uh but it's nowhere close to the IC uh packaging.
[20:08] So today we are going to focus on Silicon photonix uh you know offline we can talk about the design assembly and and the test but today we are going to talk about what are the challenges on Silicon photonics uh.
[20:22] Before we talk about uh silicon fonics there are multiple applications uh non started with the Telecom application uh then Cloud uh data transfer and lately.
[20:35] then Cloud uh data transfer and lately AIML meantime uh we have customers in
[20:38] AIML meantime uh we have customers in Liar as well as in industrial so the
[20:41] Liar as well as in industrial so the application of sleen photonix is uh
[20:44] application of sleen photonix is uh growing in multiple folds and that
[20:47] growing in multiple folds and that obviously create a good opportunity for
[20:49] obviously create a good opportunity for the osets so as you can see the market
[20:53] the osets so as you can see the market is growing and growing in multiple fold
[20:56] is growing and growing in multiple fold for uh next 10 years and that is a time
[21:00] for uh next 10 years and that is a time uh for large osets uh to bring in their
[21:04] uh for large osets uh to bring in their uh knoow from the IC packaging and then
[21:08] uh knoow from the IC packaging and then continue to uh you know understand and
[21:10] continue to uh you know understand and integrate Optical assembly uh within the
[21:13] integrate Optical assembly uh within the same facility and and as osets are
[21:17] same facility and and as osets are coming in like citronics or or AC of the
[21:22] coming in like citronics or or AC of the world uh and then comes s jail where
[21:24] world uh and then comes s jail where they are doing the whole uh system level
[21:26] they are doing the whole uh system level integration so as industry grows
[21:29] integration so as industry grows obviously there are challenges uh
[21:31] obviously there are challenges uh without the uh you know if you don't
[21:33] without the uh you know if you don't resolve the challenges up front then
[21:36] resolve the challenges up front then Skilling in high volume will be very uh
[21:38] Skilling in high volume will be very uh diff difficult so talking about the
[21:40] diff difficult so talking about the challenges we have uh cassendra who is
[21:44] challenges we have uh cassendra who is leading multiple process development uh
[21:47] leading multiple process development uh across the board whether there is a
[21:49] across the board whether there is a cloud uh customers or silicon photonics
[21:52] cloud uh customers or silicon photonics or light up so she's going to talk about
[21:54] or light up so she's going to talk about the challenges and how uh selectronics
[21:57] the challenges and how uh selectronics engine team has been helping customer to
[22:00] engine team has been helping customer to mitigate these uh challenges so go ahead
[22:02] mitigate these uh challenges so go ahead cendra so when compared to standard OSAP
[22:06] cendra so when compared to standard OSAP processes think I see wirebond and I see
[22:08] processes think I see wirebond and I see flip chip the pck photonics assembly and
[22:11] flip chip the pck photonics assembly and everything that comes in with it becomes
[22:14] everything that comes in with it becomes substantially more difficult so F first
[22:16] substantially more difficult so F first of all um it's understanding that's a
[22:19] of all um it's understanding that's a newer Market there's fewer standards
[22:22] newer Market there's fewer standards there's many more things to worry about
[22:25] there's many more things to worry about um and this this really affects your
[22:27] um and this this really affects your time and your your timeline for doing
[22:30] time and your your timeline for doing everything from prototypes to NPI to doe
[22:33] everything from prototypes to NPI to doe um and these variables really do affect
[22:36] um and these variables really do affect the entire process so just from a very
[22:40] the entire process so just from a very technical perspective um first webon and
[22:43] technical perspective um first webon and flip ship generally can be considered to
[22:45] flip ship generally can be considered to have three three critical placement
[22:47] have three three critical placement critical axes rotation also matters but
[22:50] critical axes rotation also matters but it matters much less than linear
[22:52] it matters much less than linear position but adding these three more
[22:54] position but adding these three more axes sometimes with very tight tolerance
[22:57] axes sometimes with very tight tolerance is really increases exponentially the
[23:00] is really increases exponentially the challenge
[23:02] challenge assembly um then there's also the issue
[23:05] assembly um then there's also the issue that many of the current automated
[23:08] that many of the current automated assembly machines are designed to be a
[23:10] assembly machines are designed to be a custom Solution that's designed for one
[23:12] custom Solution that's designed for one product and is expected to continue
[23:15] product and is expected to continue along with that product and then on the
[23:17] along with that product and then on the other hand reconfigurable platforms tend
[23:19] other hand reconfigurable platforms tend to be more lab grade Solutions which
[23:22] to be more lab grade Solutions which aren't really suitable for doing actual
[23:25] aren't really suitable for doing actual automated production with any sort of
[23:27] automated production with any sort of repeatability this increases the time to
[23:30] repeatability this increases the time to do development does to get to production
[23:33] do development does to get to production and Optical packaging is also more
[23:35] and Optical packaging is also more highly disciplinary than the other sorts
[23:38] highly disciplinary than the other sorts of packaging that osats traditionally do
[23:41] of packaging that osats traditionally do um there's more obviously Optical
[23:44] um there's more obviously Optical engineering requirement but there's also
[23:46] engineering requirement but there's also algorithmic engineering a lot of these
[23:49] algorithmic engineering a lot of these processes that used to be very Vision
[23:51] processes that used to be very Vision alignment based and you could kind of
[23:53] alignment based and you could kind of eat the tolerance losses um now require
[23:56] eat the tolerance losses um now require more algorithmic searches and very tuned
[24:00] more algorithmic searches and very tuned in
[24:01] in profiles there's also the issue that the
[24:03] profiles there's also the issue that the materials there's just many many more of
[24:07] materials there's just many many more of them um it's partially again because
[24:09] them um it's partially again because it's an emerging industry but it makes
[24:11] it's an emerging industry but it makes it much harder for an osad to stock a
[24:13] it much harder for an osad to stock a standard bomb right so for for wire Bond
[24:16] standard bomb right so for for wire Bond you essentially have two sliders you
[24:18] you essentially have two sliders you have material and you have thickness um
[24:20] have material and you have thickness um and then you have different machines
[24:21] and then you have different machines that can make different types of bond
[24:23] that can make different types of bond heads um but for for photonics you have
[24:27] heads um but for for photonics you have you wave length you have intensity you
[24:29] you wave length you have intensity you have um Pitch type and you also have all
[24:34] have um Pitch type and you also have all of these different types of couplers you
[24:35] of these different types of couplers you know different foundaries different
[24:37] know different foundaries different applications have different couplers
[24:39] applications have different couplers couplers need different types of epoxies
[24:41] couplers need different types of epoxies sometimes you need multiple epoxies per
[24:43] sometimes you need multiple epoxies per Optical
[24:44] Optical path all of these mean that it's much
[24:47] path all of these mean that it's much harder for an oset to actually stock
[24:51] harder for an oset to actually stock stock Parts um we typically would get
[24:55] stock Parts um we typically would get people to come in and in in any other
[24:58] people to come in and in in any other type of assembly they would assume that
[24:59] type of assembly they would assume that we have material but in this one it's
[25:01] we have material but in this one it's much less likely than we actually
[25:04] much less likely than we actually do back to kind of the engineering
[25:06] do back to kind of the engineering challenges there's just much more NRE
[25:09] challenges there's just much more NRE that has to be done before photonic
[25:11] that has to be done before photonic assembly everything more or less has to
[25:14] assembly everything more or less has to be custom just given the tolerances
[25:17] be custom just given the tolerances required to be actually manipulating
[25:19] required to be actually manipulating these things in three
[25:20] these things in three dimensions um and there's also just the
[25:23] dimensions um and there's also just the reconfiguration of these machines most
[25:25] reconfiguration of these machines most wire Bond jobs or most flip chip jobs
[25:27] wire Bond jobs or most flip chip jobs can be done with machine as is photonics
[25:30] can be done with machine as is photonics jobs mostly can't there's a lot of
[25:32] jobs mostly can't there's a lot of vision alignment there's a lot of curing
[25:34] vision alignment there's a lot of curing that has to be done in that and there's
[25:37] that has to be done in that and there's obviously still the process and there's
[25:38] obviously still the process and there's optimization within
[25:40] optimization within it so we've had a few we have a few
[25:43] it so we've had a few we have a few Solutions and mitigating factors the
[25:45] Solutions and mitigating factors the biggest one is going to your oap before
[25:47] biggest one is going to your oap before you actually tape out um and kind of
[25:50] you actually tape out um and kind of doing dfm with them and you can that way
[25:53] doing dfm with them and you can that way you can resolve a lot of issues before
[25:55] you can resolve a lot of issues before they actually come up you can change
[25:56] they actually come up you can change designs or the OAC and prepare in some
[25:59] designs or the OAC and prepare in some sort of way um silitronics is also
[26:01] sort of way um silitronics is also partnering with f cont just to kind of
[26:05] partnering with f cont just to kind of unite our engineering abilities and more
[26:08] unite our engineering abilities and more on a an osat perspective and the other
[26:11] on a an osat perspective and the other thing is just that a lot of the times
[26:13] thing is just that a lot of the times recently you see these teams still come
[26:16] recently you see these teams still come to the Osa for wi Bond or flip ship and
[26:18] to the Osa for wi Bond or flip ship and they try to do Optical assembly in-house
[26:20] they try to do Optical assembly in-house using a lab grade setup but we're seeing
[26:23] using a lab grade setup but we're seeing that really reduces yield and it can
[26:25] that really reduces yield and it can increase handling damage um and
[26:28] increase handling damage um and obviously increased cycle
[26:30] obviously increased cycle times so the other thing is that there's
[26:33] times so the other thing is that there's a lot of different capabilities you need
[26:34] a lot of different capabilities you need I'll kind of burn through this because I
[26:36] I'll kind of burn through this because I already talked about but essentially
[26:37] already talked about but essentially there's a million different types of
[26:39] there's a million different types of things that need to be assembled
[26:41] things that need to be assembled together in Optics and importantly it
[26:44] together in Optics and importantly it needs to be an automatic and repeatable
[26:45] needs to be an automatic and repeatable process so you're actually hitting those
[26:47] process so you're actually hitting those yields and you're getting um good loss
[26:50] yields and you're getting um good loss characteristics I I'll hand it back to
[26:51] characteristics I I'll hand it back to Dage from more of the business angle
[26:53] Dage from more of the business angle after
[26:55] after this uh thank you Cassandra talking
[26:58] this uh thank you Cassandra talking about the challenges and some of the
[27:01] about the challenges and some of the must have in order for oset to ramp up
[27:04] must have in order for oset to ramp up in high volume I think the key message
[27:07] in high volume I think the key message on this is that dfm uh is a must
[27:10] on this is that dfm uh is a must especially before the tape out so that
[27:12] especially before the tape out so that uh uh citronics or oset can help the
[27:15] uh uh citronics or oset can help the location of the fuell the thickness of
[27:17] location of the fuell the thickness of the pck the location of the wave guides
[27:20] the pck the location of the wave guides the size of the fiber uh the active
[27:23] the size of the fiber uh the active alignment process the tooling for
[27:25] alignment process the tooling for gripper so there are so many varable
[27:28] gripper so there are so many varable uh which need to be addressed up front
[27:30] uh which need to be addressed up front uh so that uh it's easy to ramp up down
[27:33] uh so that uh it's easy to ramp up down the road for high volume here is one of
[27:36] the road for high volume here is one of the case study we have done for one of
[27:39] the case study we have done for one of our Cloud customers where you can see uh
[27:43] our Cloud customers where you can see uh that the true uh mix of the phonix IC
[27:46] that the true uh mix of the phonix IC and the regular IC along with the fiber
[27:48] and the regular IC along with the fiber array is happening and the integation is
[27:52] array is happening and the integation is happening more and more across the board
[27:54] happening more and more across the board uh we have um uh experience to help
[27:58] uh we have um uh experience to help customers from designing the substrate
[28:00] customers from designing the substrate uh fabrication pick assembly uh active
[28:03] uh fabrication pick assembly uh active alignment using F contact with the
[28:06] alignment using F contact with the multiple Faus we have done anywhere from
[28:09] multiple Faus we have done anywhere from 1 by8 to almost 1 by uh 40 FAU for
[28:14] 1 by8 to almost 1 by uh 40 FAU for active alignment uh at the same time
[28:16] active alignment uh at the same time tooling is equally important uh and in
[28:18] tooling is equally important uh and in order to have all this uh process
[28:21] order to have all this uh process development uh oset needs a facility uh
[28:24] development uh oset needs a facility uh where you have everything under a clean
[28:27] where you have everything under a clean room uh more importantly having the both
[28:31] room uh more importantly having the both the I and the pck assembly Under One
[28:33] the I and the pck assembly Under One Roof uh is also the IP production and
[28:36] Roof uh is also the IP production and the collaboration between the customers
[28:39] the collaboration between the customers and the oset uh for us uh we uh because
[28:43] and the oset uh for us uh we uh because the applications uh is so much uh varies
[28:47] the applications uh is so much uh varies from Silicon photonix to Cloud uh and
[28:50] from Silicon photonix to Cloud uh and customers also have lot of hands-on
[28:53] customers also have lot of hands-on experience uh to uh while we're talking
[28:56] experience uh to uh while we're talking about the photon assembly so we allow
[28:59] about the photon assembly so we allow customers Engineers to come uh to our
[29:01] customers Engineers to come uh to our facility and develop the process
[29:04] facility and develop the process together uh side by side uh as we have
[29:07] together uh side by side uh as we have multiple uh fire contact systems in
[29:10] multiple uh fire contact systems in house and one of the reason is the IP
[29:12] house and one of the reason is the IP protection so that each customers uh are
[29:15] protection so that each customers uh are uh having a dedicated machine and once
[29:17] uh having a dedicated machine and once you have the full uh TR key uh Under One
[29:21] you have the full uh TR key uh Under One Roof uh then definitely that helps to
[29:23] Roof uh then definitely that helps to reduce the uh cycle time and that will
[29:26] reduce the uh cycle time and that will help to to ramp up for the
[29:29] help to to ramp up for the asset um and my last slide for today is
[29:33] asset um and my last slide for today is uh the market for S and photonics is is
[29:35] uh the market for S and photonics is is growing and as the market grows uh oets
[29:39] growing and as the market grows uh oets are definitely uh finding their Niche
[29:42] are definitely uh finding their Niche and finding the role how to uh go uh
[29:45] and finding the role how to uh go uh with millions of units per month uh
[29:48] with millions of units per month uh eventually hopefully it will go to
[29:50] eventually hopefully it will go to billions unit uh per month and and but
[29:53] billions unit uh per month and and but before that I think it's equally
[29:54] before that I think it's equally important to have the right o set
[29:57] important to have the right o set Partners where they have uh all the
[30:01] Partners where they have uh all the capabilities and that will reduce the
[30:03] capabilities and that will reduce the development time for the customer as
[30:05] development time for the customer as well as uh you know ramp up for the oen
[30:08] well as uh you know ramp up for the oen models and at the same time silicon
[30:11] models and at the same time silicon photonic industry and application are
[30:13] photonic industry and application are growing uh definitely there need to be
[30:15] growing uh definitely there need to be more uh active alinement system there
[30:17] more uh active alinement system there need to be more wable testing there need
[30:20] need to be more wable testing there need to be more incoming uh whether it's a 3D
[30:23] to be more incoming uh whether it's a 3D x-ray or doing the inspection 200x of
[30:26] x-ray or doing the inspection 200x of the pick on the side so lot of
[30:28] the pick on the side so lot of opportunities for uh not only for the
[30:30] opportunities for uh not only for the oset but also the equipment suppliers uh
[30:33] oset but also the equipment suppliers uh and uh as a industry mature for
[30:35] and uh as a industry mature for photonics the investment will be
[30:37] photonics the investment will be continued needed from selectronics point
[30:40] continued needed from selectronics point of view we have made several investment
[30:43] of view we have made several investment uh which uh helping us to establish oset
[30:46] uh which uh helping us to establish oset uh in terms of USA and these investment
[30:50] uh in terms of USA and these investment uh have definitely want the confidence
[30:53] uh have definitely want the confidence uh from our customers and as we are
[30:55] uh from our customers and as we are helping our customers our learning uh
[30:58] helping our customers our learning uh curve is getting uh uh wider and and and
[31:02] curve is getting uh uh wider and and and obviously deeper and as a as a result uh
[31:06] obviously deeper and as a as a result uh we have been winning Confidence from
[31:07] we have been winning Confidence from hyperscalers tier one t OEM uh as well
[31:11] hyperscalers tier one t OEM uh as well as me many leading startups and with
[31:14] as me many leading startups and with these startups to support as I said IP
[31:16] these startups to support as I said IP production and cod development in
[31:18] production and cod development in selectronics is equally important uh so
[31:21] selectronics is equally important uh so that's how we see the oset uh uh
[31:24] that's how we see the oset uh uh business model and ecosystem is
[31:26] business model and ecosystem is developing for next few years um and
[31:30] developing for next few years um and definitely uh we are here to support and
[31:33] definitely uh we are here to support and uh help our customers as much as we can
[31:36] uh help our customers as much as we can uh obviously for oset having everything
[31:38] uh obviously for oset having everything one roof or under uh you know one
[31:41] one roof or under uh you know one purchase order as you say is we
[31:43] purchase order as you say is we important so thank you for the
[31:44] important so thank you for the opportunity and uh thank you H thank you
[31:48] opportunity and uh thank you H thank you very much Das let me remind everyone for
[31:50] very much Das let me remind everyone for all the questions you have for the
[31:52] all the questions you have for the rashan Cassandra please write them using
[31:54] rashan Cassandra please write them using the Q&A button exactly at 1122 Pacific
[31:58] the Q&A button exactly at 1122 Pacific time we are going to go through all the
[32:00] time we are going to go through all the questions and make sure you get the
[32:02] questions and make sure you get the answers that you need but before that
[32:04] answers that you need but before that dur just one quick question for you many
[32:07] dur just one quick question for you many of your customers all your customers
[32:09] of your customers all your customers demand that all the chips package chips
[32:11] demand that all the chips package chips that they get from you are properly
[32:13] that they get from you are properly tested to what extent is testing still a
[32:17] tested to what extent is testing still a challenge when it comes to the
[32:19] challenge when it comes to the osat yes yes that's a very good question
[32:22] osat yes yes that's a very good question uh when it comes to the IC packaging the
[32:24] uh when it comes to the IC packaging the testing is very straightforward you just
[32:26] testing is very straightforward you just do open short test for optical uh
[32:30] do open short test for optical uh testing uh you know it's still uh a
[32:32] testing uh you know it's still uh a growing growing application lot of
[32:35] growing growing application lot of customers are doing uh testing inhouse
[32:37] customers are doing uh testing inhouse because there's lot of Ip involved
[32:39] because there's lot of Ip involved recently we start working uh and
[32:41] recently we start working uh and bringing in more testing uh equipments
[32:44] bringing in more testing uh equipments by working with the quantify so that we
[32:47] by working with the quantify so that we can start supporting customers inhouse
[32:50] can start supporting customers inhouse to shorten the true you know the process
[32:52] to shorten the true you know the process development time so yeah we we customers
[32:55] development time so yeah we we customers are asking us more and more but as you
[32:56] are asking us more and more but as you said
[32:57] said they are still uh uh you know they want
[33:00] they are still uh uh you know they want to protect their IP but uh you know for
[33:02] to protect their IP but uh you know for the startups they have to start using
[33:05] the startups they have to start using oset so that they can do the true uh
[33:08] oset so that they can do the true uh Trend key down the road yeah but that's
[33:09] Trend key down the road yeah but that's a great question thank you much Das for
[33:11] a great question thank you much Das for the companies United States for many
[33:13] the companies United States for many years you have been working with osats
[33:16] years you have been working with osats in Thailand like FAO Focus or in Europe
[33:19] in Thailand like FAO Focus or in Europe like like fix you have silitronics in
[33:21] like like fix you have silitronics in United States allowing you to keep the
[33:23] United States allowing you to keep the supply chain and the manufacturing in
[33:25] supply chain and the manufacturing in the country the is great to have you on
[33:28] the country the is great to have you on board let's continue but now let's go
[33:31] board let's continue but now let's go back to The Foundry because I hope the
[33:33] back to The Foundry because I hope the connection now is stable enough and we
[33:35] connection now is stable enough and we can finish your presentation
[33:38] can finish your presentation understanding The Foundry view on the
[33:41] understanding The Foundry view on the Silicon photonics for AI vas let's try
[33:44] Silicon photonics for AI vas let's try again the floor the attention of
[33:45] again the floor the attention of everyone goes back to you okay thank you
[33:47] everyone goes back to you okay thank you I apologize I lost internet I'm going to
[33:49] I apologize I lost internet I'm going to turn off my video because I'm on my
[33:51] turn off my video because I'm on my phone hopefully yall can hear me and and
[33:54] phone hopefully yall can hear me and and I'm going to
[33:55] I'm going to present start FSE
[33:59] present start FSE yeah sure okay can can everybody see yes
[34:04] yeah sure okay can can everybody see yes please go to presentation mode yeah yeah
[34:06] please go to presentation mode yeah yeah and display settings
[34:08] and display settings swap okay I apologize uh so so as as as
[34:12] swap okay I apologize uh so so as as as I was mentioning uh this presentation is
[34:14] I was mentioning uh this presentation is going to be talking about The Foundry
[34:17] going to be talking about The Foundry and the
[34:18] and the ecosystem uh what I'm going to start off
[34:20] ecosystem uh what I'm going to start off with just giving you a brief overview of
[34:22] with just giving you a brief overview of GF photonics as a technology and then I
[34:25] GF photonics as a technology and then I want to say a few words on the ecosystem
[34:27] want to say a few words on the ecosystem the work that Global fundies as a
[34:29] the work that Global fundies as a company is doing but then also you know
[34:31] company is doing but then also you know in the industry demands uh with respect
[34:34] in the industry demands uh with respect to the ecosystem so let me just start
[34:36] to the ecosystem so let me just start from the beginning uh GF offers a
[34:39] from the beginning uh GF offers a silicon photonics technology uh that is
[34:41] silicon photonics technology uh that is sort of keyed towards you know
[34:43] sort of keyed towards you know generative AI as these these generative
[34:45] generative AI as these these generative AI systems are becoming more and more
[34:47] AI systems are becoming more and more complex so is the need for sort of low
[34:50] complex so is the need for sort of low loss uh high-speed communication energy
[34:53] loss uh high-speed communication energy efficient communication between uh the
[34:56] efficient communication between uh the compute clusters
[34:57] compute clusters and towards that GF offers this GF
[35:01] and towards that GF offers this GF photonics technology what it does you
[35:03] photonics technology what it does you can think of it as a sort of a system on
[35:05] can think of it as a sort of a system on a chip uh where we integrate you know
[35:08] a chip uh where we integrate you know the electronics at least the analog
[35:10] the electronics at least the analog Electronics with with the photonic so
[35:13] Electronics with with the photonic so you can see in the stop sort of
[35:14] you can see in the stop sort of cross-section on the left uh I have
[35:17] cross-section on the left uh I have represented the electronics with the
[35:19] represented the electronics with the nfat and and the Pat but then in
[35:22] nfat and and the Pat but then in addition to that there is a very rich
[35:24] addition to that there is a very rich feat feature Rich sort of offering with
[35:27] feat feature Rich sort of offering with respect to photonic components which
[35:29] respect to photonic components which includes Active Components such as the
[35:31] includes Active Components such as the modulator and detector on the right
[35:34] modulator and detector on the right there are passive components such as you
[35:36] there are passive components such as you know multiple options in terms of wave
[35:37] know multiple options in terms of wave guide silicon and silicon nitrate and
[35:41] guide silicon and silicon nitrate and and we Al also offer high efficiency
[35:44] and we Al also offer high efficiency heaters uh with with the sealed undercut
[35:47] heaters uh with with the sealed undercut in addition to this PO lighten light out
[35:50] in addition to this PO lighten light out a plan of record is this vgroup based
[35:54] a plan of record is this vgroup based fiber connector and uh
[35:57] fiber connector and uh polite in uh we are also working towards
[36:00] polite in uh we are also working towards an on die laser now all of this is
[36:03] an on die laser now all of this is integrated on a 300 mm wafer now in
[36:07] integrated on a 300 mm wafer now in addition to sort of the technology piece
[36:09] addition to sort of the technology piece of things uh We've also paid a lot of
[36:11] of things uh We've also paid a lot of attention towards the packaging aspect
[36:14] attention towards the packaging aspect so uh we are working towards copper
[36:17] so uh we are working towards copper pillars and tsvs and then and I will go
[36:20] pillars and tsvs and then and I will go into some of the details uh later on in
[36:22] into some of the details uh later on in the
[36:23] the presentation in addition to that uh from
[36:26] presentation in addition to that uh from a test ility perspective at least for
[36:28] a test ility perspective at least for the component level test the device
[36:30] the component level test the device level test we also have wafer level test
[36:33] level test we also have wafer level test capability now the beauty or at least
[36:36] capability now the beauty or at least the approach that GF has taken with
[36:37] the approach that GF has taken with respect to Silicon photonics is that we
[36:39] respect to Silicon photonics is that we want to create a platform a platform
[36:42] want to create a platform a platform that is flexible it's feature and it
[36:45] that is flexible it's feature and it allows for multiple architectures to
[36:48] allows for multiple architectures to coexist so you could have different
[36:50] coexist so you could have different customers coming in with different AR
[36:52] customers coming in with different AR circuit architectures or system
[36:54] circuit architectures or system architectures and we are able to support
[36:56] architectures and we are able to support that with this GF photonics platform so
[36:59] that with this GF photonics platform so on the left I'm showing you on on this
[37:01] on the left I'm showing you on on this on this slide I'm showing you all just
[37:02] on this slide I'm showing you all just two sort of
[37:04] two sort of architectural uh options on the left is
[37:07] architectural uh options on the left is the fast and narrow which is the
[37:08] the fast and narrow which is the electrical switching uh scaling and on
[37:11] electrical switching uh scaling and on the right which is the optical scaling
[37:14] the right which is the optical scaling and what that means is in terms of fast
[37:16] and what that means is in terms of fast and narrow you can think about a narrow
[37:18] and narrow you can think about a narrow stream of electrical data which is
[37:20] stream of electrical data which is modulated using a ring modulator and the
[37:23] modulated using a ring modulator and the electrical signal is imprinted on the
[37:26] electrical signal is imprinted on the laser output and output through uh a
[37:28] laser output and output through uh a fiber at 100 gbits per second now on the
[37:32] fiber at 100 gbits per second now on the slow and wide side you can think of it
[37:34] slow and wide side you can think of it as you know a parallel stream of you
[37:37] as you know a parallel stream of you know lower speed electrical signals
[37:40] know lower speed electrical signals coming in these electrical signals are
[37:43] coming in these electrical signals are moxed and imprinted on multiple
[37:46] moxed and imprinted on multiple wavelengths of light and and they are
[37:48] wavelengths of light and and they are multiplexed into a fiber to give you
[37:51] multiplexed into a fiber to give you sort of the same high-speed output now
[37:54] sort of the same high-speed output now the beauty of GF photonics as a
[37:56] the beauty of GF photonics as a technology is that both of these
[37:59] technology is that both of these architectures can be made using the the
[38:02] architectures can be made using the the the standard pdk that GF offers and
[38:05] the standard pdk that GF offers and there's enough flexibility built into
[38:07] there's enough flexibility built into the process to allow for both of these
[38:09] the process to allow for both of these architectures so with that I want to
[38:11] architectures so with that I want to sort of just uh change tax and start
[38:14] sort of just uh change tax and start talking a bit about the work the GF
[38:16] talking a bit about the work the GF Photon GF is doing in terms of the
[38:19] Photon GF is doing in terms of the ecosystem so just as a definition of the
[38:21] ecosystem so just as a definition of the ecosystem it it requires you know it's a
[38:23] ecosystem it it requires you know it's a complicated system it requires multiple
[38:27] complicated system it requires multiple uh uh companies to work with each other
[38:29] uh uh companies to work with each other each company has a specialty they they
[38:33] each company has a specialty they they uh they are focused on a particular part
[38:36] uh they are focused on a particular part of the ecosystem but there's a high
[38:38] of the ecosystem but there's a high level of
[38:39] level of interdependency uh between these in uh
[38:42] interdependency uh between these in uh components so if you think of it
[38:44] components so if you think of it starting with the fabulous uh companies
[38:47] starting with the fabulous uh companies in the middle we we we need edas Ed
[38:50] in the middle we we we need edas Ed tools to be able to design components a
[38:52] tools to be able to design components a Foundry to be able to
[38:54] Foundry to be able to produce and and then I've also added in
[38:57] produce and and then I've also added in the sort of electronic photonic IP
[38:59] the sort of electronic photonic IP ecosystem which is very nent for silicon
[39:02] ecosystem which is very nent for silicon photonics but is also starting to grow
[39:05] photonics but is also starting to grow in addition we just heard from an osad
[39:08] in addition we just heard from an osad partner uh we're going to be hearing
[39:09] partner uh we're going to be hearing from test and then there's you know the
[39:12] from test and then there's you know the suppliers who provide things like lasers
[39:14] suppliers who provide things like lasers fibers interposers and things like that
[39:17] fibers interposers and things like that so all of these all of these sort of
[39:20] so all of these all of these sort of pieces of the ecosystem have to come
[39:22] pieces of the ecosystem have to come together in in in sort of getting
[39:25] together in in in sort of getting silicon photonic to fulfill sort of the
[39:28] silicon photonic to fulfill sort of the the full benefit uh of the technology so
[39:32] the full benefit uh of the technology so starting off with the Eda ecosystem uh
[39:35] starting off with the Eda ecosystem uh Global fundies as a company has been
[39:36] Global fundies as a company has been working with with sort of the leading or
[39:39] working with with sort of the leading or the name brand uh Eda tool vendors
[39:42] the name brand uh Eda tool vendors starting with Anis Cadence seens and
[39:44] starting with Anis Cadence seens and synopsis there are certain components
[39:46] synopsis there are certain components that Global fund provides such as the P
[39:49] that Global fund provides such as the P Sals the design roles and and uh the
[39:52] Sals the design roles and and uh the models but all of the other
[39:55] models but all of the other infrastructure to allow our customers to
[39:57] infrastructure to allow our customers to design and get things first time right
[40:00] design and get things first time right are provided by these Eda vendors and it
[40:03] are provided by these Eda vendors and it is important for these Eda vendors to
[40:06] is important for these Eda vendors to get the tools right to allow for the
[40:08] get the tools right to allow for the ease of design and and the flexibility
[40:12] ease of design and and the flexibility that that customers need and as silicon
[40:14] that that customers need and as silicon photonics develops there will be more
[40:17] photonics develops there will be more and more integration of these circuits
[40:19] and more integration of these circuits on a single chip and so the complexity
[40:22] on a single chip and so the complexity of these tools is expected to grow along
[40:25] of these tools is expected to grow along with this this technology so this is
[40:27] with this this technology so this is just a view of you know uh what GF
[40:30] just a view of you know uh what GF offers as as a company there's there's a
[40:33] offers as as a company there's there's a there's a very large featur library that
[40:36] there's a very large featur library that we offer of passive and active
[40:39] we offer of passive and active components with the intention of a
[40:42] components with the intention of a customers getting a first you know first
[40:44] customers getting a first you know first time right but then also the ease of
[40:47] time right but then also the ease of Designing uh to this to this technology
[40:50] Designing uh to this to this technology and and photonics does bring you know
[40:54] and and photonics does bring you know sort of differences from from an camos
[40:57] sort of differences from from an camos and electrical pdk in that you have curv
[41:00] and electrical pdk in that you have curv linear shapes and there are a lot of
[41:02] linear shapes and there are a lot of second ordered effects that were not
[41:04] second ordered effects that were not present in electronics that need to be
[41:07] present in electronics that need to be accounted for such as sort of the
[41:09] accounted for such as sort of the nonlinear behavior of components you
[41:12] nonlinear behavior of components you know under high power
[41:13] know under high power lasers all of these are of course
[41:15] lasers all of these are of course comprehended as part of the pdk and
[41:18] comprehended as part of the pdk and we're working very closely with the Eda
[41:20] we're working very closely with the Eda tool vendors to make sure that all of
[41:23] tool vendors to make sure that all of these sort of features are captured in
[41:25] these sort of features are captured in the pdk we uh offer now I I I found this
[41:29] the pdk we uh offer now I I I found this sort of slide uh searching the internet
[41:31] sort of slide uh searching the internet it is from Cadence but it also gives
[41:34] it is from Cadence but it also gives sort of a good segue into you know where
[41:37] sort of a good segue into you know where where where does Eda stop and where does
[41:39] where where does Eda stop and where does packaging start and what it's showing is
[41:42] packaging start and what it's showing is the complexity of design as these these
[41:46] the complexity of design as these these Technologies are progressing while this
[41:48] Technologies are progressing while this is talking about more of semas but I'll
[41:50] is talking about more of semas but I'll put it in in in context of the photonic
[41:52] put it in in in context of the photonic side so you can see with every
[41:54] side so you can see with every generation as things have progressed on
[41:57] generation as things have progressed on the SEMA side there is more and more uh
[42:01] the SEMA side there is more and more uh simulation and and look ahead and Tool
[42:04] simulation and and look ahead and Tool complexity that has come into play as
[42:07] complexity that has come into play as part of the Eda ecosystem and you can
[42:09] part of the Eda ecosystem and you can see this is sort of where we are right
[42:11] see this is sort of where we are right now it talks about hetrogeneous systems
[42:14] now it talks about hetrogeneous systems and and now you need to start
[42:15] and and now you need to start comprehending system level functional
[42:17] comprehending system level functional timing power reliability thermal
[42:20] timing power reliability thermal mechanical and the case of silicon
[42:22] mechanical and the case of silicon photonics also Optical performance and
[42:25] photonics also Optical performance and this is where things start getting
[42:27] this is where things start getting complex uh because Eda needs to not only
[42:30] complex uh because Eda needs to not only comprehend the circuit but also start
[42:32] comprehend the circuit but also start comprehending the system systems now
[42:36] comprehending the system systems now what I've drawn here is sort of just
[42:39] what I've drawn here is sort of just just to give a flavor of the complexity
[42:41] just to give a flavor of the complexity of the packaging uh ecosystem that has
[42:44] of the packaging uh ecosystem that has come into play uh over for silicon
[42:47] come into play uh over for silicon photonic so what I'm showing you is sort
[42:50] photonic so what I'm showing you is sort of three forms of packaging one on the
[42:53] of three forms of packaging one on the Silicon Bridge one on a silicon interpos
[42:56] Silicon Bridge one on a silicon interpos and one on a multi-chip module an
[42:59] and one on a multi-chip module an organic laminate and once you start
[43:01] organic laminate and once you start considering the options of these
[43:03] considering the options of these packaging uh sort of choices and the
[43:06] packaging uh sort of choices and the corresponding components packaging
[43:09] corresponding components packaging components that go into it so for
[43:11] components that go into it so for example uh there are multiple solutions
[43:13] example uh there are multiple solutions for fiber attach once you start thinking
[43:16] for fiber attach once you start thinking about silicon Bridge versus interposer
[43:18] about silicon Bridge versus interposer you may have mixed pitch versus very you
[43:21] you may have mixed pitch versus very you know very minute bumps uh you need to
[43:24] know very minute bumps uh you need to account for the electrical IC attached
[43:27] account for the electrical IC attached to the optical engine and of course uh
[43:31] to the optical engine and of course uh things like thermal components heat
[43:33] things like thermal components heat sinks whether whether the components are
[43:35] sinks whether whether the components are lided or UNL so this sort of this is
[43:39] lided or UNL so this sort of this is sort of the ecosystem that we are
[43:42] sort of the ecosystem that we are operating in from a silicon photonics
[43:45] operating in from a silicon photonics perspective and one of the challenge is
[43:47] perspective and one of the challenge is is depending upon the solution whether
[43:49] is depending upon the solution whether you approach it sort of Bottoms Up or
[43:51] you approach it sort of Bottoms Up or tops down you could reach very different
[43:54] tops down you could reach very different sort of solutions if you start so if you
[43:56] sort of solutions if you start so if you look at the ecosystem starting from the
[43:58] look at the ecosystem starting from the electronic or the photonic circuit uh
[44:02] electronic or the photonic circuit uh going up you know through the packaging
[44:04] going up you know through the packaging to the module build to a system build
[44:06] to the module build to a system build there are certain criteria that come
[44:08] there are certain criteria that come into play whether you do it tops down or
[44:10] into play whether you do it tops down or Bottoms Up starting from tops down you
[44:13] Bottoms Up starting from tops down you have to think about power consumption
[44:15] have to think about power consumption signal Integrity you know thermal
[44:17] signal Integrity you know thermal profiles where you need a heat sink
[44:19] profiles where you need a heat sink cooling plate the type of cooling that
[44:21] cooling plate the type of cooling that you require whether it's liquid or air
[44:23] you require whether it's liquid or air cooled and and you know reach of fibers
[44:27] cooled and and you know reach of fibers Etc but then if you look at it from
[44:29] Etc but then if you look at it from bottoms up uh you know you need to think
[44:33] bottoms up uh you know you need to think about the the bump pitch whether it's
[44:35] about the the bump pitch whether it's mix pitch or or fixed pitch the type of
[44:38] mix pitch or or fixed pitch the type of fiber attach that is required a a and so
[44:41] fiber attach that is required a a and so one of the challenges in the industry
[44:43] one of the challenges in the industry which the industry will you know soon
[44:44] which the industry will you know soon need to deal with is the fact that
[44:47] need to deal with is the fact that depending upon its Bottoms Up approach
[44:49] depending upon its Bottoms Up approach and top down approach uh there could be
[44:52] and top down approach uh there could be sort of a very large Matrix of solutions
[44:55] sort of a very large Matrix of solutions and and that's the reason I'm glad the
[44:56] and and that's the reason I'm glad the APC has sort of Taken on this this
[44:58] APC has sort of Taken on this this Challenge on of getting us all together
[45:01] Challenge on of getting us all together as sort of an ecosystem to talk about
[45:03] as sort of an ecosystem to talk about these challenges so from a GF
[45:06] these challenges so from a GF perspective I would say you know there
[45:07] perspective I would say you know there are certain components that we are sort
[45:10] are certain components that we are sort of prefetching as we know as we talk to
[45:12] of prefetching as we know as we talk to customers and and if you start looking
[45:15] customers and and if you start looking through sort of the unit processes
[45:17] through sort of the unit processes whether it is copper pillar copper
[45:18] whether it is copper pillar copper pillar receive pads tsvs through the
[45:21] pillar receive pads tsvs through the copper pillar with copper pillar for
[45:23] copper pillar with copper pillar for sort of the Next Generation 3D Packaging
[45:26] sort of the Next Generation 3D Packaging passive fiber attached and and Laser
[45:29] passive fiber attached and and Laser attached these are components that we've
[45:31] attached these are components that we've already built into the GF photonics
[45:34] already built into the GF photonics technology and we continuing to work
[45:36] technology and we continuing to work with our osat partners in in order to
[45:38] with our osat partners in in order to put in the infrastructure to actually go
[45:40] put in the infrastructure to actually go to high volume
[45:43] to high volume manufacturing uh the last last slide I
[45:46] manufacturing uh the last last slide I wanted to just talk about is sort of the
[45:47] wanted to just talk about is sort of the the nent sort of photonic IP ecosystem
[45:51] the nent sort of photonic IP ecosystem that is just starting to version up so
[45:54] that is just starting to version up so you can think you sort of break up a a
[45:56] you can think you sort of break up a a comp you know a a chip starting with an
[45:59] comp you know a a chip starting with an IP level uh sort of a component level
[46:03] IP level uh sort of a component level you know once you start putting multiple
[46:04] you know once you start putting multiple components together at a circuit level
[46:07] components together at a circuit level and then once you start putting circuits
[46:08] and then once you start putting circuits together at a chiplet level and I do
[46:11] together at a chiplet level and I do believe that in order to take the full
[46:14] believe that in order to take the full potential of this technology and sort of
[46:16] potential of this technology and sort of get it to a market and get it to people
[46:18] get it to a market and get it to people who are not necessarily photonics
[46:20] who are not necessarily photonics designers you will start seeing a lot
[46:23] designers you will start seeing a lot more uh of this sort of photonic IP
[46:26] more uh of this sort of photonic IP ecosystem start burgeoning uh in this
[46:30] ecosystem start burgeoning uh in this industry and so in closing I just wanted
[46:33] industry and so in closing I just wanted to sort of close with with with sort of
[46:34] to sort of close with with with sort of two bullets one is that GF photonics as
[46:37] two bullets one is that GF photonics as a technology uh from a Foundry offering
[46:40] a technology uh from a Foundry offering perspective what we've taken the
[46:42] perspective what we've taken the approach is we want to build a platform
[46:45] approach is we want to build a platform that is flexible enough to be able to
[46:47] that is flexible enough to be able to support multiple solutions for
[46:49] support multiple solutions for interconnects but then this sort of
[46:51] interconnects but then this sort of flexibility can be extended further into
[46:54] flexibility can be extended further into different markets or also so that's the
[46:56] different markets or also so that's the approach that GF GF has taken and part
[47:00] approach that GF GF has taken and part of that is also as a Foundry we need to
[47:02] of that is also as a Foundry we need to be able to scale it not only in terms of
[47:05] be able to scale it not only in terms of circuit Solutions but then also scale it
[47:08] circuit Solutions but then also scale it in terms of volumes and that's the
[47:10] in terms of volumes and that's the approach that we have taken and in order
[47:12] approach that we have taken and in order for silicon fonics as a technology uh to
[47:16] for silicon fonics as a technology uh to be sort of the vibrant and get to its
[47:18] be sort of the vibrant and get to its full potential we will need s of the
[47:21] full potential we will need s of the disparate people in the industry the
[47:23] disparate people in the industry the osad the IP ecosystem the Eda ecosystem
[47:27] osad the IP ecosystem the Eda ecosystem the test ecosystem to all work together
[47:30] the test ecosystem to all work together and perhaps I'll use the word coherently
[47:32] and perhaps I'll use the word coherently work together uh to be able to get this
[47:34] work together uh to be able to get this technology to its full
[47:36] technology to its full potential thank you thank you very much
[47:39] potential thank you thank you very much vas for a great great presentation it is
[47:42] vas for a great great presentation it is a paradig sh for us to have the GF
[47:44] a paradig sh for us to have the GF photonic pdk available for everyone and
[47:48] photonic pdk available for everyone and I will encourage everyone to come back
[47:50] I will encourage everyone to come back to you and to see how they can already
[47:52] to you and to see how they can already start using this this pdk is actually
[47:55] start using this this pdk is actually quite easy to to access I have just one
[47:57] quite easy to to access I have just one quick question for you before we go to
[47:58] quick question for you before we go to the next step so one thing that for me
[48:01] the next step so one thing that for me is the challenge for silicon photonics
[48:03] is the challenge for silicon photonics today is the wafer level testing so the
[48:06] today is the wafer level testing so the company that is following the osat can
[48:09] company that is following the osat can have a easier way to save their
[48:11] have a easier way to save their customers what is the current status on
[48:13] customers what is the current status on wafer level testing what is the current
[48:16] wafer level testing what is the current yield of silicon photonis can you
[48:17] yield of silicon photonis can you comment on that yeah so I I I I wanton
[48:19] comment on that yeah so I I I I wanton comment on yield I think the yields are
[48:21] comment on yield I think the yields are pretty high just because of the fact
[48:23] pretty high just because of the fact that photonic components in in general
[48:26] that photonic components in in general tend to be large in term in terms of
[48:28] tend to be large in term in terms of size so in terms of defect limited yield
[48:31] size so in terms of defect limited yield you know that's usually not an issue I
[48:33] you know that's usually not an issue I I'll make a comment on the wafer level
[48:35] I'll make a comment on the wafer level testing now wafer level testing is just
[48:37] testing now wafer level testing is just fundamental and GF photonics offers sort
[48:39] fundamental and GF photonics offers sort of two pieces of this sort of puzzle for
[48:41] of two pieces of this sort of puzzle for waer level testing so the first thing is
[48:44] waer level testing so the first thing is we do have as I mentioned right at the
[48:46] we do have as I mentioned right at the beginning we do have uh uh a capability
[48:49] beginning we do have uh uh a capability that was developed internally at GF to
[48:51] that was developed internally at GF to do wer level testing of components uh
[48:54] do wer level testing of components uh it's called an inline Optical test
[48:56] it's called an inline Optical test system and and and what that does it
[48:59] system and and and what that does it allows GF to be able to test components
[49:01] allows GF to be able to test components at wer level and and use the output of
[49:05] at wer level and and use the output of it to be able to create statistical
[49:07] it to be able to create statistical plots of w you know individual device
[49:10] plots of w you know individual device components so that we can make sure that
[49:12] components so that we can make sure that the customer understands that the wafer
[49:14] the customer understands that the wafer is healthy now there is also this whole
[49:16] is healthy now there is also this whole aspect of system level testing you know
[49:20] aspect of system level testing you know at speed and that is work that we are
[49:22] at speed and that is work that we are contined to work with our you know test
[49:24] contined to work with our you know test Partners on is there is a lot of
[49:27] Partners on is there is a lot of development that needs to be done not
[49:30] development that needs to be done not only from test programs but also
[49:33] only from test programs but also fixtures uh you know people are talking
[49:36] fixtures uh you know people are talking about Periscope probes to be able to
[49:38] about Periscope probes to be able to probe the edge of the wafer so that
[49:40] probe the edge of the wafer so that that's sort of that ecosystem is is
[49:43] that's sort of that ecosystem is is something that needs to be developed the
[49:45] something that needs to be developed the other point that I do want to make is
[49:47] other point that I do want to make is the fact that we integrate both
[49:48] the fact that we integrate both electronics and photonics what could
[49:50] electronics and photonics what could conceive about you know self test uh
[49:55] conceive about you know self test uh because of the fact that the electronics
[49:56] because of the fact that the electronics is built on the same die as the
[50:00] is built on the same die as the photonics what could conceive of you
[50:02] photonics what could conceive of you know self test circuits which do not
[50:04] know self test circuits which do not require an optical interface that you
[50:06] require an optical interface that you can test everything convert it into the
[50:08] can test everything convert it into the electrical signals and at least talk
[50:11] electrical signals and at least talk about self test circuits so that's sort
[50:13] about self test circuits so that's sort of the approach that that we've taken
[50:15] of the approach that that we've taken thank you very much vas I think the go
[50:17] thank you very much vas I think the go of self testing on photon is a paradigm
[50:20] of self testing on photon is a paradigm shift and we are all very very happy to
[50:22] shift and we are all very very happy to see the foundaries working on that let's
[50:24] see the foundaries working on that let's continue let's go now to the contract
[50:28] continue let's go now to the contract manufacturer does everyone in the room
[50:30] manufacturer does everyone in the room now we have several hundreds of people
[50:32] now we have several hundreds of people in the room right now does everyone in
[50:33] in the room right now does everyone in the room know that Jil has
[50:36] the room know that Jil has 260,000
[50:37] 260,000 employees this is one of the giant
[50:40] employees this is one of the giant singer industry that has been developed
[50:42] singer industry that has been developed Contra manufacturing service in many
[50:44] Contra manufacturing service in many different markets but they are taking
[50:46] different markets but they are taking silicon photonics very seriously as we
[50:49] silicon photonics very seriously as we all know of the news from Intel few
[50:52] all know of the news from Intel few months ago no offer thank you very much
[50:55] months ago no offer thank you very much V with us thank you for representing Jil
[50:57] V with us thank you for representing Jil and thank you for taking the supply
[50:58] and thank you for taking the supply chain role of Contra manufacturing the
[51:00] chain role of Contra manufacturing the floor and the attention of everyone is
[51:03] floor and the attention of everyone is yours thank you Jose and thank you
[51:05] yours thank you Jose and thank you everybody for listening in my name is
[51:08] everybody for listening in my name is nir and I'm a senior technical director
[51:10] nir and I'm a senior technical director osad and Sil fonics at jable Cloud uh
[51:14] osad and Sil fonics at jable Cloud uh I'll talk a bit about our view and
[51:16] I'll talk a bit about our view and mostly observations on volume Skilling
[51:19] mostly observations on volume Skilling of silicon photonics and the ecosystem
[51:22] of silicon photonics and the ecosystem as vas termed it uh so let's go
[51:27] as vas termed it uh so let's go ahead so just for those who are not very
[51:30] ahead so just for those who are not very familiar with jble jel has been around
[51:32] familiar with jble jel has been around for nearly 60 years as as you said Jose
[51:37] for nearly 60 years as as you said Jose more than 250,000 employees more than
[51:39] more than 250,000 employees more than 100 manufacturing sites across the globe
[51:42] 100 manufacturing sites across the globe and in fiscal year 2023 uh we had a
[51:45] and in fiscal year 2023 uh we had a revenue of 34.7 billion and just
[51:48] revenue of 34.7 billion and just recently we re-entered the S&P
[51:51] recently we re-entered the S&P 500 in terms of our Solutions and this
[51:54] 500 in terms of our Solutions and this roughly lines up jables divisions we are
[51:57] roughly lines up jables divisions we are engaged in all of these end markets and
[52:00] engaged in all of these end markets and we need to keep in mind that yes in this
[52:03] we need to keep in mind that yes in this conversation we're talking very much
[52:04] conversation we're talking very much about AI machine learning which Maps
[52:07] about AI machine learning which Maps into Cloud but actually underlying
[52:09] into Cloud but actually underlying Technologies and capabilities will
[52:11] Technologies and capabilities will develop in this field are all also may
[52:14] develop in this field are all also may be applicable to other end markets such
[52:16] be applicable to other end markets such as Automotive Healthcare
[52:24] Etc next
[52:26] Etc next slide um so specifically for focusing on
[52:30] slide um so specifically for focusing on the the AIML ja cloud has a footprint in
[52:34] the the AIML ja cloud has a footprint in the service space with white box
[52:36] the service space with white box Hardware volume deployed to hyperscalers
[52:38] Hardware volume deployed to hyperscalers and other clients required tailored
[52:40] and other clients required tailored solutions for their data center needs
[52:43] solutions for their data center needs and Ja is engage across all aspects of
[52:47] and Ja is engage across all aspects of hyperscale data centers from compute to
[52:48] hyperscale data centers from compute to networking as well as Power
[52:51] networking as well as Power Cooling and we see integrated Optics
[52:54] Cooling and we see integrated Optics being critical to a IML platforms sooner
[52:56] being critical to a IML platforms sooner rather than later in general a
[52:59] rather than later in general a IML H as our clients are telling us are
[53:04] IML H as our clients are telling us are driving a lot of Technologies which have
[53:06] driving a lot of Technologies which have existed perhaps for a while but have not
[53:08] existed perhaps for a while but have not been adopted such as uh liquid
[53:11] been adopted such as uh liquid cooling and second photonics and really
[53:14] cooling and second photonics and really Force becoming a foring function to
[53:16] Force becoming a foring function to introduce these now and Jael is now in
[53:20] introduce these now and Jael is now in getting more and more involved in silic
[53:21] getting more and more involved in silic fonic because our clients are asking us
[53:23] fonic because our clients are asking us to be in this in its business
[53:26] to be in this in its business impacting uh specifically in cyon
[53:29] impacting uh specifically in cyon photonics our clients asking us to
[53:32] photonics our clients asking us to assist with verticalization simplifying
[53:34] assist with verticalization simplifying Supply chains and also accelerate the
[53:36] Supply chains and also accelerate the technology adoption to enable high
[53:38] technology adoption to enable high volume manufacturing at the cost targets
[53:40] volume manufacturing at the cost targets that are
[53:43] that are needed so some observation on the C flux
[53:47] needed so some observation on the C flux manufacturing ecosystem and this is kind
[53:49] manufacturing ecosystem and this is kind of an example flow and there are many
[53:52] of an example flow and there are many other flows that exist but this is one
[53:54] other flows that exist but this is one for how transceiver could be built so on
[53:58] for how transceiver could be built so on the top left there's wafer Fab for
[54:00] the top left there's wafer Fab for instance Global foundaries there'll be
[54:02] instance Global foundaries there'll be wafer test some clients will not do it
[54:05] wafer test some clients will not do it at the Fab could be that they own it
[54:06] at the Fab could be that they own it domestically or internally then if this
[54:09] domestically or internally then if this is a chip that needs to be bumped to get
[54:12] is a chip that needs to be bumped to get it bumped might be something you now
[54:15] it bumped might be something you now need to find a partner that's yet a
[54:17] need to find a partner that's yet a different place then to single Aid and
[54:20] different place then to single Aid and dice might yet be another place if
[54:22] dice might yet be another place if you're doing some uh bumping and you're
[54:24] you're doing some uh bumping and you're doing flip chip assembly to create a
[54:27] doing flip chip assembly to create a multi-chip module that might be yet
[54:29] multi-chip module that might be yet another vendor or stop across the globe
[54:32] another vendor or stop across the globe and then you get to fiber attach lid
[54:36] and then you get to fiber attach lid attach uh EO test and then finally
[54:40] attach uh EO test and then finally actually build it into a transceiver
[54:41] actually build it into a transceiver model so we have this multihop supply
[54:45] model so we have this multihop supply chain and our view is that there's no
[54:48] chain and our view is that there's no One-Stop shop for everything right now
[54:52] One-Stop shop for everything right now and it's exacerbated by the multitude of
[54:54] and it's exacerbated by the multitude of existing Packaging schemes that are
[54:56] existing Packaging schemes that are really driven by the optical IO
[54:59] really driven by the optical IO constraints
[55:00] constraints and what this means is that essentially
[55:05] and what this means is that essentially yield cost and time to manufacture are
[55:10] yield cost and time to manufacture are relatively high right now and our
[55:11] relatively high right now and our clients are eager to simplify this
[55:14] clients are eager to simplify this process but
[55:17] process but uh so far I think we as the industry
[55:19] uh so far I think we as the industry have not really stepped up to deliver
[55:23] have not really stepped up to deliver this one of the as I said was
[55:26] this one of the as I said was the the plethora of uh of packaging and
[55:31] the the plethora of uh of packaging and specifically if you look at the optical
[55:33] specifically if you look at the optical packaging I have here a few examples of
[55:35] packaging I have here a few examples of interesting Solutions uh vgrs from
[55:37] interesting Solutions uh vgrs from Global foundaries uh nubis
[55:39] Global foundaries uh nubis Communications with a vertical solution
[55:43] Communications with a vertical solution and detachable connectors from teram
[55:45] and detachable connectors from teram Mount but there are many more uh
[55:48] Mount but there are many more uh Solutions right now in the market H
[55:51] Solutions right now in the market H being product
[55:53] being product productized and
[55:55] productized and the problem is when we have all these
[55:56] the problem is when we have all these different Technologies and no
[55:57] different Technologies and no standardization is that it makes it hard
[55:59] standardization is that it makes it hard for CMS to commit ahead uh in terms of
[56:03] for CMS to commit ahead uh in terms of uh Capital
[56:05] uh Capital Investments
[56:08] Investments so this not just the fiber attach
[56:10] so this not just the fiber attach machine itself but rather that these
[56:12] machine itself but rather that these different Optical IO schemes also impact
[56:16] different Optical IO schemes also impact the rest of the processing and package
[56:17] the rest of the processing and package design on the electrical side
[56:20] design on the electrical side so the optical side really dictates a
[56:23] so the optical side really dictates a lot of the package flow for everything
[56:25] lot of the package flow for everything else and the fragility of it ER drives
[56:29] else and the fragility of it ER drives the sequences especially with Optical
[56:32] the sequences especially with Optical epoxies and fibers allowing you limited
[56:34] epoxies and fibers allowing you limited the thermal budget and the fiber
[56:36] the thermal budget and the fiber handling complicating uh as well
[56:40] handling complicating uh as well Automation and
[56:45] handling so in different slides there's
[56:48] handling so in different slides there's always this um exponential growth and in
[56:52] always this um exponential growth and in different years if you look at the
[56:53] different years if you look at the Historical forecast
[56:55] Historical forecast it was always imminent this exponential
[56:59] it was always imminent this exponential growth uh where we'll transition to a
[57:01] growth uh where we'll transition to a high volume
[57:04] high volume regime
[57:05] regime uh in the past uh large OAS have been
[57:09] uh in the past uh large OAS have been burned with Investments that did not
[57:12] burned with Investments that did not translate into volumes as this hockey
[57:14] translate into volumes as this hockey stick Behavior did not materialize at
[57:16] stick Behavior did not materialize at the timing that was
[57:18] the timing that was forecasted but we believe that the
[57:20] forecasted but we believe that the inflection point is near and currently
[57:23] inflection point is near and currently held back more by the manufacturing
[57:25] held back more by the manufacturing ecosystem which has not geared itself to
[57:27] ecosystem which has not geared itself to be able to deliver at the speed at Cost
[57:30] be able to deliver at the speed at Cost targets rather than market
[57:32] targets rather than market demand but there is a chicken and egg
[57:35] demand but there is a chicken and egg type of situation we're facing that
[57:37] type of situation we're facing that volumes volume orders are waiting for
[57:39] volumes volume orders are waiting for the appropriate supply chain solution
[57:41] the appropriate supply chain solution that can deliver at the cost and volume
[57:43] that can deliver at the cost and volume targets but building up the Capac the
[57:46] targets but building up the Capac the capacity requires a significant
[57:49] capacity requires a significant investment upfront which is at high risk
[57:52] investment upfront which is at high risk if it doesn't
[57:53] if it doesn't materialize so so and it's difficult to
[57:56] materialize so so and it's difficult to do a gradual build for high volume
[57:59] do a gradual build for high volume manufacturer um as a cost tget is only
[58:02] manufacturer um as a cost tget is only realizable when you put high volume
[58:04] realizable when you put high volume tooling
[58:11] on in terms of uh another observation
[58:14] on in terms of uh another observation that I like to make is that there's this
[58:16] that I like to make is that there's this blurring of perhaps traditional uh
[58:19] blurring of perhaps traditional uh roles on the interface between
[58:22] roles on the interface between The Foundry and the o set we observe
[58:25] The Foundry and the o set we observe that because of all the custom needs of
[58:27] that because of all the custom needs of silicon phonics Founders are forced
[58:30] silicon phonics Founders are forced to hold the hands of their clients
[58:32] to hold the hands of their clients Beyond wafer
[58:34] Beyond wafer Fab into far back end of line and
[58:37] Fab into far back end of line and post-processing and in absence of
[58:40] post-processing and in absence of significant silic phic volumes large
[58:43] significant silic phic volumes large osets may not be incentivized to make
[58:46] osets may not be incentivized to make significant efforts into custom
[58:47] significant efforts into custom solutions that address silk and photonic
[58:49] solutions that address silk and photonic Wafers so that's one area of blurring of
[58:52] Wafers so that's one area of blurring of the line where uh
[58:55] the line where uh the Fab and the oset blurs on the other
[58:59] the Fab and the oset blurs on the other side because of fiber handling and
[59:03] side because of fiber handling and because of
[59:05] because of the sequencing of
[59:08] the sequencing of integration uh the line between cm and
[59:11] integration uh the line between cm and oset also is blurred on the electrical
[59:15] oset also is blurred on the electrical assembly and the optical
[59:17] assembly and the optical assembly uh so we're seeing this blur of
[59:21] assembly uh so we're seeing this blur of lines and we need to figure out as an
[59:23] lines and we need to figure out as an industry how we provide the service that
[59:26] industry how we provide the service that our clients can now scale
[59:31] with
[59:33] with jable has a track record uh in
[59:36] jable has a track record uh in integrated Optics since the acquisition
[59:38] integrated Optics since the acquisition of AOC Technologies back in 2015 and
[59:40] of AOC Technologies back in 2015 and followed with manufacturing ramp of
[59:42] followed with manufacturing ramp of electronic coherent Optical engines and
[59:44] electronic coherent Optical engines and a presence that's now expanded with the
[59:46] a presence that's now expanded with the Acquisitions of the Intel transceiver
[59:49] Acquisitions of the Intel transceiver business we think we have the scale in
[59:51] business we think we have the scale in the global footprint to chase down
[59:53] the global footprint to chase down service intimacy and and simplify Supply
[59:55] service intimacy and and simplify Supply chains we believe we can assess this
[59:59] chains we believe we can assess this standardization and participate in
[01:00:01] standardization and participate in supply chain simplification to unlock
[01:00:03] supply chain simplification to unlock the velocity of silk and photonic
[01:00:04] the velocity of silk and photonic technology
[01:00:06] technology adoption in our view Partnerships can be
[01:00:09] adoption in our view Partnerships can be a path forwards to advance the
[01:00:11] a path forwards to advance the integrated Optics uh ecosystem and we're
[01:00:14] integrated Optics uh ecosystem and we're engaged in uh some of these discussions
[01:00:17] engaged in uh some of these discussions in industry and we look to further
[01:00:19] in industry and we look to further engage uh in more discussions as this
[01:00:22] engage uh in more discussions as this topic continues to evolve so
[01:00:25] topic continues to evolve so my my final words would be that we look
[01:00:28] my my final words would be that we look forward to hearing what other players
[01:00:31] forward to hearing what other players views are and to further this discussion
[01:00:34] views are and to further this discussion at the ofc
[01:00:36] at the ofc panel thank you very much I really
[01:00:39] panel thank you very much I really appreciate when you come with your
[01:00:41] appreciate when you come with your doubts about the volume needs because of
[01:00:43] doubts about the volume needs because of course J is the manufacturer of the
[01:00:45] course J is the manufacturer of the GoPro camera it's manufacturer of many
[01:00:47] GoPro camera it's manufacturer of many components in everybody's mobile phones
[01:00:50] components in everybody's mobile phones and you are thinking now silicon Photon
[01:00:52] and you are thinking now silicon Photon we had their needs my question to you is
[01:00:54] we had their needs my question to you is because honestly I have no doubt that we
[01:00:56] because honestly I have no doubt that we will have demands on volume production
[01:00:58] will have demands on volume production of silicon photonics my question to you
[01:01:00] of silicon photonics my question to you is everything is everything else ready
[01:01:03] is everything is everything else ready on the contract manufacturing I mean the
[01:01:05] on the contract manufacturing I mean the micro lenses the chips the electronic
[01:01:08] micro lenses the chips the electronic chips the the packaging the volume
[01:01:11] chips the the packaging the volume production the equipment necessary is
[01:01:13] production the equipment necessary is everything else ready should the big
[01:01:15] everything else ready should the big customers should the consumer companies
[01:01:17] customers should the consumer companies come to you with demands of silicon
[01:01:20] come to you with demands of silicon photonics well the the consumers are
[01:01:22] photonics well the the consumers are coming us with demands uh uh our end
[01:01:26] coming us with demands uh uh our end clients uh but there's a still a process
[01:01:30] clients uh but there's a still a process to establish the volume and jabble has
[01:01:34] to establish the volume and jabble has already a presence in volume now with
[01:01:37] already a presence in volume now with silic photonics and in other products
[01:01:40] silic photonics and in other products but yeah there's more work to be done
[01:01:43] but yeah there's more work to be done and there are areas in the intersections
[01:01:47] and there are areas in the intersections between osat and CM that we need to
[01:01:50] between osat and CM that we need to figure out exactly how we do it so and
[01:01:55] figure out exactly how we do it so and and to do it to do it you need
[01:01:56] and to do it to do it you need automation to do the volume production
[01:01:59] automation to do the volume production of of silicon photonics you need
[01:02:02] of of silicon photonics you need equipment that can provide automation on
[01:02:04] equipment that can provide automation on the assembly packaging and test that's
[01:02:06] the assembly packaging and test that's why the next speaker is a good friend of
[01:02:08] why the next speaker is a good friend of yours and a good friend of mine we go
[01:02:10] yours and a good friend of mine we go all the way to Akim to Breman to Germany
[01:02:13] all the way to Akim to Breman to Germany to meet Thoren venam the CEO of one of
[01:02:16] to meet Thoren venam the CEO of one of the most beautiful success stories in
[01:02:19] the most beautiful success stories in the photonic industry for those of you
[01:02:20] the photonic industry for those of you who want to know the story Google D
[01:02:23] who want to know the story Google D varen camp host poo and read the article
[01:02:26] varen camp host poo and read the article it's a beautiful story but now we are
[01:02:28] it's a beautiful story but now we are talking about what F cont can do for the
[01:02:30] talking about what F cont can do for the contract manufacturer and the osats the
[01:02:32] contract manufacturer and the osats the floor and the attention goes to you
[01:02:34] floor and the attention goes to you Thoren thank you you can see my
[01:02:37] Thoren thank you you can see my presentation yeah um Gorilla Glass
[01:02:39] presentation yeah um Gorilla Glass clear the uh I would like to thank you
[01:02:42] clear the uh I would like to thank you uh the the previous speaker for for this
[01:02:44] uh the the previous speaker for for this excellent talks and and a lot of points
[01:02:47] excellent talks and and a lot of points were were raised by global foundies C
[01:02:49] were were raised by global foundies C tonics and Jab what I will might tackle
[01:02:52] tonics and Jab what I will might tackle now in my presentation my name is to
[01:02:55] now in my presentation my name is to pankam I'm CEO and one of the founders
[01:02:58] pankam I'm CEO and one of the founders of Ponch Ponch stands for fiber
[01:03:01] of Ponch Ponch stands for fiber connecting technology so that is one of
[01:03:03] connecting technology so that is one of the key ingredients of the topic we are
[01:03:05] the key ingredients of the topic we are talking about uh um and uh even though
[01:03:10] talking about uh um and uh even though we talking about today about the AI
[01:03:12] we talking about today about the AI Revolution um the uh the topic uh is is
[01:03:17] Revolution um the uh the topic uh is is much broader than than just AI so
[01:03:20] much broader than than just AI so photonics today is is everywhere and as
[01:03:22] photonics today is is everywhere and as we heard in in ja uh we see the same
[01:03:25] we heard in in ja uh we see the same things it's not only Telecom Datacom
[01:03:27] things it's not only Telecom Datacom it's it's ligher it's biosensing it's
[01:03:29] it's it's ligher it's biosensing it's Quantum Technologies much much more
[01:03:31] Quantum Technologies much much more areas where fonics today is is key uh
[01:03:35] areas where fonics today is is key uh for uh for the future and a lot of these
[01:03:38] for uh for the future and a lot of these areas uh have pics inside today so um
[01:03:44] areas uh have pics inside today so um and we as an equipment manufacturer um
[01:03:47] and we as an equipment manufacturer um where I will give the view uh today are
[01:03:50] where I will give the view uh today are are seeing these pic now implemented in
[01:03:52] are seeing these pic now implemented in all these uh all these devices um um
[01:03:57] all these uh all these devices um um very challenging um
[01:04:00] very challenging um uh components and and and uh structures
[01:04:05] uh components and and and uh structures we see um uh all different types um no
[01:04:10] we see um uh all different types um no standardization and that is what we have
[01:04:12] standardization and that is what we have to work and live with as a as an
[01:04:15] to work and live with as a as an equipment manufacturer if you see the
[01:04:18] equipment manufacturer if you see the product assembly and I'm just touching
[01:04:20] product assembly and I'm just touching the areas we see there's many more uh
[01:04:23] the areas we see there's many more uh areas the whole electronic assembly what
[01:04:25] areas the whole electronic assembly what you have to tackle um uh we have to to
[01:04:28] you have to tackle um uh we have to to start with a with a known good die uh in
[01:04:31] start with a with a known good die uh in in the components um the silic the
[01:04:34] in the components um the silic the integrated photonic uh devices so we are
[01:04:38] integrated photonic uh devices so we are need um and the previous speakers have
[01:04:40] need um and the previous speakers have touched that we are in need for
[01:04:42] touched that we are in need for electrooptical uh weer level testing
[01:04:45] electrooptical uh weer level testing so we we need to integrate uh the Laser
[01:04:48] so we we need to integrate uh the Laser Source as we heard previously uh so we
[01:04:52] Source as we heard previously uh so we need High precision die bonding systems
[01:04:55] need High precision die bonding systems which bring the integrated uh which
[01:04:57] which bring the integrated uh which brings the laser into the integrated
[01:04:58] brings the laser into the integrated circuit uh we need fiber array and
[01:05:02] circuit uh we need fiber array and assembly and Alignment tools and
[01:05:04] assembly and Alignment tools and equipment which is producing uh the
[01:05:07] equipment which is producing uh the fiber uh studs themselves so we need uh
[01:05:11] fiber uh studs themselves so we need uh fiber interconnect technology and
[01:05:13] fiber interconnect technology and finally in order to really enable our um
[01:05:17] finally in order to really enable our um volume production we need reliable
[01:05:20] volume production we need reliable machines which are potentially also work
[01:05:23] machines which are potentially also work in line as is today totally common in in
[01:05:26] in line as is today totally common in in electronics assembly so um I will touch
[01:05:30] electronics assembly so um I will touch these topics and I will start with the
[01:05:32] these topics and I will start with the wafer level testing and I will all touch
[01:05:34] wafer level testing and I will all touch this very briefly because about every uh
[01:05:37] this very briefly because about every uh item you can give a whole speech itself
[01:05:40] item you can give a whole speech itself so if I look at the electrooptical wer
[01:05:43] so if I look at the electrooptical wer level testing um we have uh two main uh
[01:05:48] level testing um we have uh two main uh topics we need to tackle so there's
[01:05:49] topics we need to tackle so there's grating coupling uh we have to uh we
[01:05:52] grating coupling uh we have to uh we have to enable or we have Edge coupling
[01:05:55] have to enable or we have Edge coupling devices on wafer level um which we have
[01:05:58] devices on wafer level um which we have to as equipment man manufacturer we have
[01:06:01] to as equipment man manufacturer we have to um we have to generate for this Edge
[01:06:05] to um we have to generate for this Edge coupling specifically and it was also
[01:06:06] coupling specifically and it was also raised before um there's this little uh
[01:06:09] raised before um there's this little uh periscopes um which were developed by um
[01:06:14] periscopes um which were developed by um a company called uh Vanguard
[01:06:16] a company called uh Vanguard automation these units we are using in
[01:06:19] automation these units we are using in our wafer testers and these little
[01:06:21] our wafer testers and these little periscopes are super small um highly
[01:06:24] periscopes are super small um highly efficient and you can print them onto
[01:06:26] efficient and you can print them onto the fiber array and you actually dive
[01:06:28] the fiber array and you actually dive them into the wafer in the trenches and
[01:06:31] them into the wafer in the trenches and do Edge cing so perfect working
[01:06:33] do Edge cing so perfect working technology we uh build a lot of graer
[01:06:35] technology we uh build a lot of graer level test systems which have this
[01:06:37] level test systems which have this capability and um it's working fully
[01:06:40] capability and um it's working fully automatic um and also very short
[01:06:43] automatic um and also very short coupling time so that is the technology
[01:06:45] coupling time so that is the technology which we uh see a lot um right now
[01:06:48] which we uh see a lot um right now implemented into the industry um this is
[01:06:51] implemented into the industry um this is such a machine how it looks like today
[01:06:53] such a machine how it looks like today uh one of the tools we
[01:06:57] uh one of the tools we developed um so generally there's a
[01:06:59] developed um so generally there's a wafer table which is usually 300 mm
[01:07:01] wafer table which is usually 300 mm heatable um and coolable um then there
[01:07:06] heatable um and coolable um then there is uh the optical alignment engines so
[01:07:09] is uh the optical alignment engines so there usually two Sixaxis alignment
[01:07:11] there usually two Sixaxis alignment engines inside which then come from the
[01:07:13] engines inside which then come from the top or from the bottom uh that is the
[01:07:15] top or from the bottom uh that is the new system we're currently develop so
[01:07:17] new system we're currently develop so electrical probing from the top
[01:07:19] electrical probing from the top electrical Optical probing from the
[01:07:21] electrical Optical probing from the bottom so there's um technology which is
[01:07:24] bottom so there's um technology which is Dev developed for different type of uh
[01:07:27] Dev developed for different type of uh devices but there's more much more than
[01:07:30] devices but there's more much more than just the wafer table and the uh um the
[01:07:34] just the wafer table and the uh um the optical um interconnects which need to
[01:07:37] optical um interconnects which need to be taken care of there's for sure the
[01:07:39] be taken care of there's for sure the high rut um requirements we need to see
[01:07:42] high rut um requirements we need to see there's short alignment times and and uh
[01:07:44] there's short alignment times and and uh the customers want to have the each die
[01:07:46] the customers want to have the each die 100% tested uh in in very short time
[01:07:50] 100% tested uh in in very short time then we need to have interaction with um
[01:07:52] then we need to have interaction with um with the instrumentation uh company so
[01:07:56] with the instrumentation uh company so like qualify phonics um has the next uh
[01:07:59] like qualify phonics um has the next uh the next talk so we have actually the
[01:08:03] the next talk so we have actually the whole range of of Quantified phonic um
[01:08:07] whole range of of Quantified phonic um products already as part of our as part
[01:08:10] products already as part of our as part of our libraries and our machines
[01:08:13] of our libraries and our machines there's RF requirements we have to
[01:08:14] there's RF requirements we have to tackle so up to 65 uh gahz um which
[01:08:19] tackle so up to 65 uh gahz um which which we need to tackle the whole data
[01:08:20] which we need to tackle the whole data handle handling needs to be done
[01:08:22] handle handling needs to be done automatically so um you can you need to
[01:08:25] automatically so um you can you need to be operate this this machines as slave
[01:08:28] be operate this this machines as slave or as a master so sometimes the optical
[01:08:30] or as a master so sometimes the optical test equipment is is driving the machine
[01:08:32] test equipment is is driving the machine so all these things we need to take care
[01:08:34] so all these things we need to take care of that there's much more to say about
[01:08:37] of that there's much more to say about wer level testing um but there's as I
[01:08:41] wer level testing um but there's as I mentioned already there's a lot of
[01:08:43] mentioned already there's a lot of assembly requirements we need to we need
[01:08:45] assembly requirements we need to we need to tackle as a as an equipment
[01:08:47] to tackle as a as an equipment manufacturer so there's a is a need for
[01:08:49] manufacturer so there's a is a need for high Precision die bonding and what we
[01:08:52] high Precision die bonding and what we develop for that is pretty much U an
[01:08:56] develop for that is pretty much U an infrared view viewing system where we
[01:08:59] infrared view viewing system where we can look through the Silicon uh
[01:09:02] can look through the Silicon uh wafer and we look from the bottoms
[01:09:05] wafer and we look from the bottoms through that device uh using uh infrared
[01:09:09] through that device uh using uh infrared um uh light and infrared
[01:09:12] um uh light and infrared cameras and that allows us to see if I
[01:09:16] cameras and that allows us to see if I have a die above the uh the Silicon
[01:09:20] have a die above the uh the Silicon wafer um I can see from the bottom side
[01:09:23] wafer um I can see from the bottom side uh the alignments structures on the Chip
[01:09:26] uh the alignments structures on the Chip And while doing that um I can do the
[01:09:29] And while doing that um I can do the alignment so I can do an XY manipulation
[01:09:32] alignment so I can do an XY manipulation um can see the D above the above the um
[01:09:37] um can see the D above the above the um the wafer and finally I can combine that
[01:09:40] the wafer and finally I can combine that with a with a laser bonding process so I
[01:09:43] with a with a laser bonding process so I have a little video how that can uh and
[01:09:46] have a little video how that can uh and how it's actually looking
[01:09:48] how it's actually looking um so you see the The Little D on the on
[01:09:53] um so you see the The Little D on the on the pickup tool you have the optical
[01:09:54] the pickup tool you have the optical component or the Silicon device on the
[01:09:57] component or the Silicon device on the bottom side you bring the Dy above the
[01:10:01] bottom side you bring the Dy above the Silicon component um you see the solder
[01:10:04] Silicon component um you see the solder in
[01:10:04] in green now I switch to the IR camera up
[01:10:08] green now I switch to the IR camera up on the bottom side and you see the Dy
[01:10:11] on the bottom side and you see the Dy manipulated above the Silicon component
[01:10:15] manipulated above the Silicon component so you see the wave guides you see the
[01:10:17] so you see the wave guides you see the the alignment marks I bring them all
[01:10:19] the alignment marks I bring them all into alignment I bring them into um the
[01:10:23] into alignment I bring them into um the right
[01:10:26] position and finally after I've did did
[01:10:29] position and finally after I've did did the alignment I can start the bonding
[01:10:31] the alignment I can start the bonding process and I do that uh potentially
[01:10:33] process and I do that uh potentially through the same Optics or I can switch
[01:10:36] through the same Optics or I can switch Optics and have the have the laser doing
[01:10:38] Optics and have the have the laser doing the bonding process so that's our
[01:10:40] the bonding process so that's our approach to very high Precision
[01:10:42] approach to very high Precision Alignment um we have some some alignment
[01:10:45] Alignment um we have some some alignment results actually we reach um
[01:10:48] results actually we reach um approximately plus minus uh 200
[01:10:50] approximately plus minus uh 200 nanometer um 3 Sigma um align
[01:10:54] nanometer um 3 Sigma um align accuracy and for us uh that is then
[01:10:57] accuracy and for us uh that is then combined with the with a localized
[01:11:00] combined with the with a localized through silicon laser soldering process
[01:11:02] through silicon laser soldering process so shooting with a laser from the bottom
[01:11:04] so shooting with a laser from the bottom side through the Silicon wafer um to
[01:11:07] side through the Silicon wafer um to solder or to to to heat the solder and
[01:11:10] solder or to to to heat the solder and get the get the soldering done and this
[01:11:13] get the get the soldering done and this is what we are currently working on to
[01:11:15] is what we are currently working on to do this whole thing also wether level
[01:11:17] do this whole thing also wether level based um that is something which we will
[01:11:20] based um that is something which we will finish most probably early uh end of q1
[01:11:24] finish most probably early uh end of q1 or early Q2 so that's um what's cooking
[01:11:27] or early Q2 so that's um what's cooking in our lab right now so very happy about
[01:11:30] in our lab right now so very happy about the results so far but um the next step
[01:11:33] the results so far but um the next step is not only the the die needs to be
[01:11:36] is not only the the die needs to be assembled one of the most critical parts
[01:11:38] assembled one of the most critical parts and this is what uh is also manifested
[01:11:42] and this is what uh is also manifested in our name the fibers need to be uh
[01:11:44] in our name the fibers need to be uh connected so is that fiber connecting
[01:11:47] connected so is that fiber connecting Technologies that's a key part we are
[01:11:48] Technologies that's a key part we are dealing with um that is our bread and
[01:11:50] dealing with um that is our bread and butter
[01:11:51] butter business and um um that we can align
[01:11:56] business and um um that we can align fibers passively and actively um um most
[01:12:00] fibers passively and actively um um most of people um know we have 1,200 machines
[01:12:06] of people um know we have 1,200 machines in the field which are somehow related
[01:12:09] in the field which are somehow related um with with fibers and Optical assembly
[01:12:11] um with with fibers and Optical assembly but what's new is really the generation
[01:12:15] but what's new is really the generation of the whole um fiber studs so that's uh
[01:12:20] of the whole um fiber studs so that's uh what we are currently working on a full
[01:12:22] what we are currently working on a full automatic process to generate these uh
[01:12:26] automatic process to generate these uh fiber units um stripping cleaving
[01:12:29] fiber units um stripping cleaving automatically putting the fiber um
[01:12:32] automatically putting the fiber um ribbons and and singulated fibers into
[01:12:35] ribbons and and singulated fibers into the uh into the connectors so um that is
[01:12:39] the uh into the connectors so um that is one of the because it's flexible devices
[01:12:41] one of the because it's flexible devices a very challenging task um uh which need
[01:12:44] a very challenging task um uh which need to be handled fully automatic without
[01:12:47] to be handled fully automatic without any operator
[01:12:48] any operator intervention but one of the key
[01:12:50] intervention but one of the key ingredients for that is the fiber strip
[01:12:52] ingredients for that is the fiber strip uh and Clea process that's what I'm
[01:12:55] uh and Clea process that's what I'm showing here right now so it's actually
[01:12:57] showing here right now so it's actually the strip stripping of a fiber ribbon
[01:13:01] the strip stripping of a fiber ribbon and
[01:13:01] and finally uh the CLE the cleaving process
[01:13:05] finally uh the CLE the cleaving process of the
[01:13:06] of the ribbon uh which is done completely with
[01:13:08] ribbon uh which is done completely with the laser and then finally the in
[01:13:11] the laser and then finally the in implementation of the um the fiber
[01:13:13] implementation of the um the fiber ribbon into the into
[01:13:16] ribbon into the into connector and So currently we're working
[01:13:19] connector and So currently we're working on a process where we where we build the
[01:13:21] on a process where we where we build the the whole fiber stat fully autom IC um
[01:13:25] the whole fiber stat fully autom IC um and um within the next would say four or
[01:13:29] and um within the next would say four or five months I will be able to um uh to
[01:13:33] five months I will be able to um uh to show a complete uh real video of of the
[01:13:36] show a complete uh real video of of the complete machine that's uh what is right
[01:13:38] complete machine that's uh what is right now uh in assembly
[01:13:41] now uh in assembly so talking about this different
[01:13:43] so talking about this different processes all these um are singulated
[01:13:47] processes all these um are singulated machines but what's important for us and
[01:13:50] machines but what's important for us and this as I said it's very common in the
[01:13:51] this as I said it's very common in the electronics Industry to to generate
[01:13:54] electronics Industry to to generate these in in lines so and this will be
[01:13:58] these in in lines so and this will be for sure not only F cont machines there
[01:14:00] for sure not only F cont machines there will be many other um machines which
[01:14:03] will be many other um machines which will go into these in inline uh systems
[01:14:06] will go into these in inline uh systems there will be wire bonders there will be
[01:14:08] there will be wire bonders there will be um die bonders of other um other
[01:14:11] um die bonders of other um other manufacturers there will be the fiber
[01:14:13] manufacturers there will be the fiber attached systems so but this is not for
[01:14:17] attached systems so but this is not for uh for the
[01:14:19] uh for the AI uh world but for c for other
[01:14:21] AI uh world but for c for other customers we have uh we have generated
[01:14:24] customers we have uh we have generated photonic uh assembly process already in
[01:14:27] photonic uh assembly process already in in line this for example the aity sensor
[01:14:30] in line this for example the aity sensor uh here for German tier one
[01:14:33] uh here for German tier one um very uh
[01:14:36] um very uh sophisticated um uh lines with multiple
[01:14:41] sophisticated um uh lines with multiple active alignment processes with full
[01:14:43] active alignment processes with full automatic feeding systems with End of
[01:14:45] automatic feeding systems with End of Line testing inside this line with laser
[01:14:47] Line testing inside this line with laser marking at the end um where your data
[01:14:50] marking at the end um where your data Matrix code is already printed on the
[01:14:52] Matrix code is already printed on the device
[01:14:53] device and then finally the the final component
[01:14:55] and then finally the the final component goes out of the end feeder in blisters
[01:14:59] goes out of the end feeder in blisters so um and we envision that for also
[01:15:02] so um and we envision that for also transceiver transceivers in future um on
[01:15:05] transceiver transceivers in future um on the on the right side the picture you
[01:15:06] the on the right side the picture you see um uh a clean room and a mass
[01:15:09] see um uh a clean room and a mass production Mass manufacturing uh
[01:15:12] production Mass manufacturing uh contract manufacturer in Thailand
[01:15:14] contract manufacturer in Thailand there's 150 machines um in that bay U
[01:15:18] there's 150 machines um in that bay U which is like cassette to cassette
[01:15:20] which is like cassette to cassette handling that's mainly for for lens
[01:15:22] handling that's mainly for for lens attach for fiber fiber attach to the RX
[01:15:26] attach for fiber fiber attach to the RX side TX side so all these things are
[01:15:28] side TX side so all these things are completely
[01:15:29] completely automated but in this case you have uh
[01:15:32] automated but in this case you have uh um operators handling cassettes from one
[01:15:34] um operators handling cassettes from one machine to the other with cure processes
[01:15:37] machine to the other with cure processes in between um also that and I agree with
[01:15:40] in between um also that and I agree with the with the previous speakers there's a
[01:15:42] the with the previous speakers there's a lot to be done um um as a as a community
[01:15:46] lot to be done um um as a as a community in order to get that full full automatic
[01:15:48] in order to get that full full automatic and and high volume manufacturable if
[01:15:51] and and high volume manufacturable if you really Target the Quant ities we we
[01:15:53] you really Target the Quant ities we we talked about previously so um we are
[01:15:57] talked about previously so um we are working on on um machine learning uh for
[01:16:01] working on on um machine learning uh for these kind of processes when you want to
[01:16:03] these kind of processes when you want to go high volume you need to make sure
[01:16:05] go high volume you need to make sure your kpis are tracked um you track the
[01:16:08] your kpis are tracked um you track the yields these yields of this machines
[01:16:09] yields these yields of this machines they're running over 99 uh .5% uh yields
[01:16:14] they're running over 99 uh .5% uh yields we have uh high up we need to have high
[01:16:16] we have uh high up we need to have high up times we need to have um uh a
[01:16:20] up times we need to have um uh a preventive maintenance so if there's
[01:16:21] preventive maintenance so if there's something uh going
[01:16:24] something uh going into the wrong way inside this machine
[01:16:26] into the wrong way inside this machine you want to know that before the machine
[01:16:27] you want to know that before the machine goes down so there's uh AI processes
[01:16:30] goes down so there's uh AI processes monitoring certain process um steps and
[01:16:33] monitoring certain process um steps and also a process optimization is done with
[01:16:36] also a process optimization is done with uh with the help of AI today so we have
[01:16:40] uh with the help of AI today so we have a group working in our r&t department we
[01:16:42] a group working in our r&t department we have implemented the first custom these
[01:16:44] have implemented the first custom these kind of things on the first customers uh
[01:16:47] kind of things on the first customers uh customer sites there's Edge Computing as
[01:16:49] customer sites there's Edge Computing as Edge Computers inside the machine and
[01:16:52] Edge Computers inside the machine and then they report to a um so-called
[01:16:55] then they report to a um so-called revision control servers where we can
[01:16:56] revision control servers where we can control up to 200 machines um um with
[01:17:01] control up to 200 machines um um with the so-called copy exact processes that
[01:17:04] the so-called copy exact processes that when you want to roll out a new process
[01:17:05] when you want to roll out a new process parameter to the machines you can do
[01:17:07] parameter to the machines you can do that with one click you don't have to
[01:17:08] that with one click you don't have to run from machine to machine do that you
[01:17:10] run from machine to machine do that you can do that with with yeah in inside um
[01:17:15] can do that with with yeah in inside um these revision control servers and these
[01:17:17] these revision control servers and these is revision control server they will
[01:17:18] is revision control server they will roll out uh a new process to all the
[01:17:21] roll out uh a new process to all the machines so there's a lot of things
[01:17:23] machines so there's a lot of things happening which um
[01:17:25] happening which um will allow high volume manufacturers
[01:17:28] will allow high volume manufacturers manufacturing but um there's still huge
[01:17:31] manufacturing but um there's still huge amount of work to be done so we see
[01:17:34] amount of work to be done so we see there's required some solutions for
[01:17:36] there's required some solutions for wafer level testing High Precision die
[01:17:38] wafer level testing High Precision die bonding fiber interconnects and at the
[01:17:40] bonding fiber interconnects and at the end we have want to bring to bring all
[01:17:42] end we have want to bring to bring all of this together in in in in lines um so
[01:17:46] of this together in in in in lines um so that's what we are working on and as a
[01:17:49] that's what we are working on and as a conclusion so assembly and test are are
[01:17:53] conclusion so assembly and test are are and will remain cost drivers for the
[01:17:56] and will remain cost drivers for the phonic products um we have automated
[01:17:59] phonic products um we have automated assembly steps for most of the process
[01:18:01] assembly steps for most of the process steps necessary um and the last decade
[01:18:06] steps necessary um and the last decade was pretty much doing this in individual
[01:18:09] was pretty much doing this in individual machines when we see the next decade now
[01:18:11] machines when we see the next decade now coming um to do that is a very high
[01:18:14] coming um to do that is a very high level of automation as the previous
[01:18:16] level of automation as the previous speakers said um that's that's a key and
[01:18:19] speakers said um that's that's a key and uh um the the necessary steps in order
[01:18:22] uh um the the necessary steps in order to really
[01:18:23] to really Mass man
[01:18:24] Mass man generate high volume
[01:18:27] generate high volume manufacturing thank you very much thank
[01:18:29] manufacturing thank you very much thank you very much tan thank you for a very
[01:18:31] you very much tan thank you for a very inspiring presentation and for continue
[01:18:33] inspiring presentation and for continue being this access story on the
[01:18:35] being this access story on the automation for manufacturing and test
[01:18:37] automation for manufacturing and test you made a very strong impact on the
[01:18:40] you made a very strong impact on the assembly but at the end on the last
[01:18:41] assembly but at the end on the last slide you said both assembly and test
[01:18:44] slide you said both assembly and test Remain the need when it comes to
[01:18:46] Remain the need when it comes to automation for the test a Ware level
[01:18:48] automation for the test a Ware level testing what keeps you awake at
[01:18:51] testing what keeps you awake at night
[01:18:54] night yelling customers
[01:18:56] yelling customers no the no no what keeps me awake at
[01:18:59] no the no no what keeps me awake at night is is um really the the
[01:19:04] night is is um really the the uh the large variation of components we
[01:19:07] uh the large variation of components we see today so there there's there's no
[01:19:09] see today so there there's there's no standards every single thing we see
[01:19:12] standards every single thing we see every single uh customer comes to us
[01:19:14] every single uh customer comes to us every CM talks to us it looks always
[01:19:17] every CM talks to us it looks always different and and this is something
[01:19:19] different and and this is something which I mean finally um I always say
[01:19:21] which I mean finally um I always say that we we live from Custom ation but it
[01:19:24] that we we live from Custom ation but it also makes the process of of being
[01:19:27] also makes the process of of being compatible to high volume manufacturing
[01:19:30] compatible to high volume manufacturing very difficult you know what keeps me
[01:19:32] very difficult you know what keeps me awake at night is that for me there is
[01:19:35] awake at night is that for me there is always whenever we discuss this people
[01:19:36] always whenever we discuss this people say there is a difference between test
[01:19:39] say there is a difference between test and
[01:19:40] and characterization and I believe that
[01:19:42] characterization and I believe that currently customers require volume
[01:19:46] currently customers require volume characterization Rec volume testing they
[01:19:48] characterization Rec volume testing they really want to know the performance of
[01:19:50] really want to know the performance of every modulator of every integrated
[01:19:52] every modulator of every integrated laser of every Integrity detector down
[01:19:54] laser of every Integrity detector down to the quantum efficiency and for that
[01:19:57] to the quantum efficiency and for that making this volume characterization is
[01:19:59] making this volume characterization is the next big thing so Thursday I'm going
[01:20:01] the next big thing so Thursday I'm going to introduce now one of your partners
[01:20:03] to introduce now one of your partners the first time I met the next speaker
[01:20:05] the first time I met the next speaker the first time I met the next company
[01:20:06] the first time I met the next company quantify photonics I met them because
[01:20:08] quantify photonics I met them because their equipment was at a f contact
[01:20:11] their equipment was at a f contact machine and I go I fell in love with
[01:20:14] machine and I go I fell in love with what they can do case propstra thank you
[01:20:16] what they can do case propstra thank you very much of being our next speaker
[01:20:17] very much of being our next speaker today tell us how quantifi photonics
[01:20:20] today tell us how quantifi photonics works with ficon tech to satis the need
[01:20:23] works with ficon tech to satis the need of J Global foundies and any other
[01:20:26] of J Global foundies and any other company in the ecosystem the floor and
[01:20:27] company in the ecosystem the floor and the attention of everyone goes to you
[01:20:31] the attention of everyone goes to you all right good uh good day everybody let
[01:20:34] all right good uh good day everybody let me make sure that I can everybody see
[01:20:37] me make sure that I can everybody see my slides crystal clear all right
[01:20:41] my slides crystal clear all right perfect so thank you very much so so
[01:20:43] perfect so thank you very much so so first of all I would like to thank
[01:20:45] first of all I would like to thank Optica and and APC for for organizing
[01:20:49] Optica and and APC for for organizing this this event um you know this event
[01:20:53] this this event um you know this event really provides a unique perspective a
[01:20:56] really provides a unique perspective a very broad ecosystem perspective of of
[01:20:59] very broad ecosystem perspective of of multiple partners across the ecosystem
[01:21:01] multiple partners across the ecosystem from from Foundry all the way to end
[01:21:04] from from Foundry all the way to end user and everything in between and and
[01:21:06] user and everything in between and and and why is that so important Norm showed
[01:21:08] and why is that so important Norm showed a very nice graph with the question mark
[01:21:11] a very nice graph with the question mark so where where is that inflection point
[01:21:13] so where where is that inflection point I truly believe that
[01:21:16] I truly believe that 2024 represents this inflection point
[01:21:19] 2024 represents this inflection point you know the the Technologies are pretty
[01:21:22] you know the the Technologies are pretty mature
[01:21:23] mature but um we still need to scale them to
[01:21:26] but um we still need to scale them to high volumes and everybody is now
[01:21:28] high volumes and everybody is now preparing to be able to scale these
[01:21:31] preparing to be able to scale these integrated phonics Technologies to high
[01:21:33] integrated phonics Technologies to high volume so so that's I think is is is is
[01:21:36] volume so so that's I think is is is is very
[01:21:37] very important um so if if you look at our
[01:21:40] important um so if if you look at our industry obviously uh AI is really
[01:21:42] industry obviously uh AI is really driving a lot of this you know we've
[01:21:44] driving a lot of this you know we've seen a lot of plots where you compare
[01:21:47] seen a lot of plots where you compare the data requirements for um you know
[01:21:51] the data requirements for um you know for AI you compare it to to Moors law
[01:21:53] for AI you compare it to to Moors law and you see there's a there's a huge
[01:21:55] and you see there's a there's a huge disconnect so you know all this data
[01:21:58] disconnect so you know all this data that is required for for AI and
[01:22:01] that is required for for AI and artificial intelligence move the data
[01:22:04] artificial intelligence move the data around you need to have these very high
[01:22:07] around you need to have these very high density High Channel count hrut Optical
[01:22:11] density High Channel count hrut Optical interconnector are based on on
[01:22:13] interconnector are based on on integrated photonics and silicon
[01:22:15] integrated photonics and silicon photonics and and to my presentation
[01:22:17] photonics and and to my presentation will really focus on the testing
[01:22:19] will really focus on the testing perspective on scaling these
[01:22:21] perspective on scaling these Technologies to to high
[01:22:24] Technologies to to high volumes so what is really the problem
[01:22:27] volumes so what is really the problem statement and we've seen it you know you
[01:22:28] statement and we've seen it you know you you start with a kind of a wafer based
[01:22:32] you start with a kind of a wafer based you know product and and you need to
[01:22:34] you know product and and you need to then D it cut it into multiple
[01:22:37] then D it cut it into multiple individual dieses then you start kind of
[01:22:40] individual dieses then you start kind of the assembly process you combine it with
[01:22:42] the assembly process you combine it with the fiber array other components kind of
[01:22:44] the fiber array other components kind of create an optical subassembly and then
[01:22:47] create an optical subassembly and then you combine it with with PCB electronics
[01:22:50] you combine it with with PCB electronics and so on into a a a a module a kind of
[01:22:53] and so on into a a a a module a kind of a package product and of course you go
[01:22:55] a package product and of course you go through the product life cycle where you
[01:22:57] through the product life cycle where you start in R&D then you have the
[01:23:00] start in R&D then you have the validation and and Manufacturing and the
[01:23:02] validation and and Manufacturing and the test requirements are really going to
[01:23:04] test requirements are really going to change throughout those those Cycles so
[01:23:06] change throughout those those Cycles so it's really important to to to to to U
[01:23:10] it's really important to to to to to U pay attention to that and and to uh to
[01:23:12] pay attention to that and and to uh to develop the right kind of testing
[01:23:14] develop the right kind of testing strategy to be able to do that
[01:23:16] strategy to be able to do that successfully so if you look at these
[01:23:20] successfully so if you look at these Optical interconnects you know they
[01:23:21] Optical interconnects you know they really consist of of of a bunch of
[01:23:23] really consist of of of a bunch of photonic devices so you have lasers
[01:23:26] photonic devices so you have lasers modulators wdm components on one side
[01:23:29] modulators wdm components on one side then you have the wave guides the
[01:23:31] then you have the wave guides the receivers the grading couplers and and
[01:23:33] receivers the grading couplers and and combined that creates the you know the
[01:23:36] combined that creates the you know the the the highspeed optical interconnects
[01:23:38] the the highspeed optical interconnects but of course these different photonic
[01:23:41] but of course these different photonic devices need different methods to
[01:23:43] devices need different methods to characterize it and to determine whether
[01:23:45] characterize it and to determine whether they are they're working uh properly or
[01:23:48] they are they're working uh properly or not so that that's the The Challenge on
[01:23:49] not so that that's the The Challenge on one side so a lot of different devices
[01:23:51] one side so a lot of different devices lot of different
[01:23:53] lot of different characterization methods and then if you
[01:23:56] characterization methods and then if you look at the the measurement overviews
[01:23:57] look at the the measurement overviews you have you have a bunch of different
[01:23:59] you have you have a bunch of different things which are really you know quite
[01:24:01] things which are really you know quite quite unique on one side you have just
[01:24:03] quite unique on one side you have just pure passive measurements you know you
[01:24:05] pure passive measurements you know you just do kind of a polarization
[01:24:07] just do kind of a polarization wavelength scan of a of a of a passive
[01:24:09] wavelength scan of a of a of a passive wave guide or just a wdm component to
[01:24:13] wave guide or just a wdm component to see how it behaves but then you have
[01:24:16] see how it behaves but then you have active test you know you you power the
[01:24:18] active test you know you you power the device you control the device you you
[01:24:21] device you control the device you you provide some kind of a bias
[01:24:23] provide some kind of a bias uh for instance you know you power up
[01:24:24] uh for instance you know you power up the laser diode what does the spectrum
[01:24:26] the laser diode what does the spectrum look like is the the side mode
[01:24:28] look like is the the side mode suppression ratio is it is it sufficient
[01:24:31] suppression ratio is it is it sufficient is it meeting the spec what kind of mod
[01:24:33] is it meeting the spec what kind of mod if you look at the modulation depth of
[01:24:35] if you look at the modulation depth of your of your modulator you know you want
[01:24:37] your of your modulator you know you want to characterize the sensitivity of your
[01:24:40] to characterize the sensitivity of your receiver so those are active devices a
[01:24:41] receiver so those are active devices a little bit more complex you know you
[01:24:42] little bit more complex you know you need to power you need to hook up to
[01:24:44] need to power you need to hook up to your device and then you have the the
[01:24:46] your device and then you have the the highspeed measurement so that really
[01:24:50] highspeed measurement so that really gives you the the full bandwidth correct
[01:24:52] gives you the the full bandwidth correct characterization of your device really
[01:24:54] characterization of your device really for its in intended
[01:24:56] for its in intended highspeed uh you know operation mode so
[01:24:59] highspeed uh you know operation mode so it's like IE diagrams transmitter
[01:25:01] it's like IE diagrams transmitter characterization bit error rates or
[01:25:04] characterization bit error rates or receiver sensitivity or or even more uh
[01:25:07] receiver sensitivity or or even more uh you know exciting is to to do a a full s
[01:25:10] you know exciting is to to do a a full s parameter characterization to see what
[01:25:13] parameter characterization to see what kind of bandwidth does your transmitter
[01:25:15] kind of bandwidth does your transmitter your receiver or the whole system have
[01:25:17] your receiver or the whole system have so obviously those different categories
[01:25:19] so obviously those different categories will make it a little bit more more
[01:25:21] will make it a little bit more more complex
[01:25:23] complex so luckily we have a lot lot of
[01:25:26] so luckily we have a lot lot of solutions for that so if you look at the
[01:25:28] solutions for that so if you look at the the wide variety of instruments
[01:25:30] the wide variety of instruments available today uh you know you can you
[01:25:33] available today uh you know you can you can break it down in in categories like
[01:25:34] can break it down in in categories like you have on the input side really
[01:25:37] you have on the input side really controlling the light you have laser
[01:25:39] controlling the light you have laser polarization controller different types
[01:25:42] polarization controller different types to to bring the light into the device
[01:25:43] to to bring the light into the device then on the other side you have um you
[01:25:46] then on the other side you have um you know characterization instruments to to
[01:25:49] know characterization instruments to to to measure the light whether it's power
[01:25:51] to measure the light whether it's power levels op Optical Spectrum or bring it
[01:25:54] levels op Optical Spectrum or bring it in the electrical domain or do kind of
[01:25:56] in the electrical domain or do kind of oscilloscope measurements or kind of
[01:25:58] oscilloscope measurements or kind of photocurrent uh meter and then you have
[01:26:02] photocurrent uh meter and then you have uh measurement tools that are really
[01:26:04] uh measurement tools that are really having an input and an output like for
[01:26:06] having an input and an output like for instance a light wave component analyzer
[01:26:09] instance a light wave component analyzer which is basically an optical back to
[01:26:11] which is basically an optical back to network analyzer so you can really
[01:26:13] network analyzer so you can really measure the bandwidth of your device
[01:26:15] measure the bandwidth of your device under test or you have a bird driving a
[01:26:18] under test or you have a bird driving a high-speed data signal into the device
[01:26:20] high-speed data signal into the device and see what the response is and have a
[01:26:22] and see what the response is and have a lot of support modules to kind of
[01:26:25] lot of support modules to kind of optimize the condition so Optical
[01:26:26] optimize the condition so Optical amplifiers provide really high levels of
[01:26:29] amplifiers provide really high levels of optical signals if you need it Optical
[01:26:32] optical signals if you need it Optical uh attenuators to to Really control the
[01:26:35] uh attenuators to to Really control the power level to do like let's say kind of
[01:26:37] power level to do like let's say kind of receiver sensitivity measurements
[01:26:39] receiver sensitivity measurements switches Splitters but also very
[01:26:41] switches Splitters but also very importantly you know standard equipment
[01:26:43] importantly you know standard equipment like smus to measure and drive the
[01:26:46] like smus to measure and drive the device on the test or or electrical
[01:26:50] device on the test or or electrical Vector signal analyzers to do kind of
[01:26:52] Vector signal analyzers to do kind of Spectrum analysis and things like that
[01:26:54] Spectrum analysis and things like that so you know it's it's a it's a very
[01:26:56] so you know it's it's a it's a very complex uh you know um view on on on how
[01:27:00] complex uh you know um view on on on how to characterize the devices but but the
[01:27:02] to characterize the devices but but the solutions are there and now we need to
[01:27:04] solutions are there and now we need to just really strategize on how to do that
[01:27:08] just really strategize on how to do that so now the question is what are you
[01:27:11] so now the question is what are you going to test and when are you going to
[01:27:13] going to test and when are you going to test it so we again we go through this
[01:27:14] test it so we again we go through this this cycle from from wafer to die
[01:27:17] this cycle from from wafer to die assembly uh and module where are you
[01:27:20] assembly uh and module where are you going to test so um obviously if you
[01:27:22] going to test so um obviously if you wait too long you're going to be
[01:27:24] wait too long you're going to be potentially uh wasting resources and
[01:27:27] potentially uh wasting resources and material so you probably want to test a
[01:27:29] material so you probably want to test a little bit earlier making sure that you
[01:27:32] little bit earlier making sure that you have a known good D and especially with
[01:27:35] have a known good D and especially with let's say the the state of the industry
[01:27:37] let's say the the state of the industry you know it's pretty mature but but the
[01:27:39] you know it's pretty mature but but the yields are are still relatively low so
[01:27:42] yields are are still relatively low so you probably want to test a little bit
[01:27:43] you probably want to test a little bit more earlier in the process so that
[01:27:45] more earlier in the process so that maybe at the wafer level you you
[01:27:48] maybe at the wafer level you you determine this is my known good ey
[01:27:50] determine this is my known good ey before you commit to to all the uh the
[01:27:53] before you commit to to all the uh the next
[01:27:54] next stages so going to into a little bit
[01:27:58] stages so going to into a little bit more detail so waiver level testing so
[01:28:00] more detail so waiver level testing so that's very important state of of this
[01:28:03] that's very important state of of this of this process and a lot of people are
[01:28:06] of this process and a lot of people are asking questions about it so the of
[01:28:08] asking questions about it so the of course what you do is you you this is a
[01:28:10] course what you do is you you this is a pretty complex setup you know you have a
[01:28:12] pretty complex setup you know you have a wafer Handler you need to have a probing
[01:28:14] wafer Handler you need to have a probing system Optical as well as electrical so
[01:28:17] system Optical as well as electrical so you inject the the the signal into the
[01:28:19] you inject the the the signal into the wafer you power and activate the deves
[01:28:22] wafer you power and activate the deves you measure the optical signal and you
[01:28:24] you measure the optical signal and you you probably want to do a lot of
[01:28:25] you probably want to do a lot of characterization so um One Challenge
[01:28:29] characterization so um One Challenge here is and we probably talk a little
[01:28:31] here is and we probably talk a little bit more about it you know how how do
[01:28:32] bit more about it you know how how do you connect to the wafer optically
[01:28:35] you connect to the wafer optically there's a lot of different ways there's
[01:28:36] there's a lot of different ways there's not too much standardization but you
[01:28:38] not too much standardization but you know I think we're making a lot of good
[01:28:40] know I think we're making a lot of good progress a lot of very viable Solutions
[01:28:43] progress a lot of very viable Solutions can do this fast without too much active
[01:28:46] can do this fast without too much active alignment um but really what what we're
[01:28:49] alignment um but really what what we're hearing from the industry that at this
[01:28:51] hearing from the industry that at this stage you know a lot of folks want to do
[01:28:54] stage you know a lot of folks want to do very detailed characterization including
[01:28:57] very detailed characterization including some of the high-speed measurements
[01:28:59] some of the high-speed measurements really to make sure that you know you
[01:29:01] really to make sure that you know you select the die as non Good Die to
[01:29:05] select the die as non Good Die to basically advance in the in the
[01:29:07] basically advance in the in the manufacturing process and in in the
[01:29:09] manufacturing process and in in the first phase kind the pre-production the
[01:29:11] first phase kind the pre-production the first phase of the ramp you probably
[01:29:13] first phase of the ramp you probably want to do as much as possible hopefully
[01:29:15] want to do as much as possible hopefully then when you really go to high volume
[01:29:17] then when you really go to high volume you know you have you have more data so
[01:29:19] you know you have you have more data so that you can say like well we we can
[01:29:22] that you can say like well we we can reduce the test coverage and also maybe
[01:29:24] reduce the test coverage and also maybe reduce the percentage of devices tested
[01:29:27] reduce the percentage of devices tested but you really need to have the data so
[01:29:29] but you really need to have the data so really you know frontload the testing in
[01:29:32] really you know frontload the testing in your product life cycle in your
[01:29:34] your product life cycle in your manufacturing stages based on which you
[01:29:36] manufacturing stages based on which you can make decisions to at some point when
[01:29:39] can make decisions to at some point when you go to really high volumes to reduce
[01:29:41] you go to really high volumes to reduce the amount of testing so then the next
[01:29:45] the amount of testing so then the next step is really you have the known good
[01:29:47] step is really you have the known good di from the previous States now you need
[01:29:49] di from the previous States now you need to do the fiber array alignment and
[01:29:50] to do the fiber array alignment and maybe some some additional assembly
[01:29:52] maybe some some additional assembly steps now this is hopefully relatively
[01:29:57] steps now this is hopefully relatively high yield relatively low low risk so
[01:30:01] high yield relatively low low risk so you do the alignment you want to make
[01:30:02] you do the alignment you want to make sure that you have the optimum alignment
[01:30:04] sure that you have the optimum alignment you of course you test that and then you
[01:30:06] you of course you test that and then you know maybe as as to be on on the safe
[01:30:09] know maybe as as to be on on the safe side you want to validate that all the
[01:30:11] side you want to validate that all the channels are still working so maybe a
[01:30:13] channels are still working so maybe a quick uh scan that all the channels are
[01:30:16] quick uh scan that all the channels are still alive um because you know that the
[01:30:19] still alive um because you know that the device is working well from the previous
[01:30:20] device is working well from the previous St um and and that would be you know
[01:30:22] St um and and that would be you know kind of a a reasonable justification to
[01:30:25] kind of a a reasonable justification to limit the amount of testing however if
[01:30:27] limit the amount of testing however if the assembly steps are a little bit
[01:30:29] the assembly steps are a little bit higher risk you may want to add some
[01:30:31] higher risk you may want to add some additional testing to make sure that you
[01:30:33] additional testing to make sure that you didn't break the uh you know the device
[01:30:35] didn't break the uh you know the device on the test in in this
[01:30:37] on the test in in this stage and then if you look at the the
[01:30:40] stage and then if you look at the the kind of the final stat of the
[01:30:42] kind of the final stat of the manufacturing cycle so you you take the
[01:30:44] manufacturing cycle so you you take the optical serly you combine it with the
[01:30:47] optical serly you combine it with the PCB some the electrical IC and maybe
[01:30:49] PCB some the electrical IC and maybe some other things into kind of the form
[01:30:52] some other things into kind of the form factor the desired form factor um no
[01:30:56] factor the desired form factor um no it's really critical here this is the
[01:30:57] it's really critical here this is the last stage before you send it out to the
[01:30:59] last stage before you send it out to the end user so you really need to make sure
[01:31:01] end user so you really need to make sure that it meets all the performance
[01:31:03] that it meets all the performance requirements so you really want to test
[01:31:05] requirements so you really want to test that initially probably even over
[01:31:08] that initially probably even over temperature just to make sure that it
[01:31:09] temperature just to make sure that it really meets all the requirements and
[01:31:11] really meets all the requirements and then again once you collect more data
[01:31:13] then again once you collect more data you have a little bit of a guard Bend
[01:31:15] you have a little bit of a guard Bend for the termal behavior you can scale
[01:31:17] for the termal behavior you can scale down the testing um to to to minimize it
[01:31:20] down the testing um to to to minimize it but still there's a lot of testing
[01:31:21] but still there's a lot of testing involved and it seems that um you know
[01:31:25] involved and it seems that um you know you can optimize the testing but um
[01:31:28] you can optimize the testing but um there's probably still a fair amount of
[01:31:30] there's probably still a fair amount of 100% testing required you know like
[01:31:33] 100% testing required you know like every channel you need to make sure that
[01:31:34] every channel you need to make sure that it is working at at high speed so it
[01:31:37] it is working at at high speed so it seems right now that it's still
[01:31:39] seems right now that it's still desirable to do highspeed testing for
[01:31:41] desirable to do highspeed testing for all the channels that means that there
[01:31:43] all the channels that means that there is really a need to parallelize the
[01:31:45] is really a need to parallelize the testing as much as possible to be have
[01:31:48] testing as much as possible to be have an Optimum test flow and Manufacturing
[01:31:51] an Optimum test flow and Manufacturing flow to be able to test these devices
[01:31:53] flow to be able to test these devices and make sure that they are 100% working
[01:31:55] and make sure that they are 100% working before they go being deployed uh in the
[01:31:59] before they go being deployed uh in the field so in conclusion so you know what
[01:32:03] field so in conclusion so you know what what is a successful testing strategy
[01:32:06] what is a successful testing strategy you know you need to make sure that you
[01:32:07] you know you need to make sure that you have all these test functions to be able
[01:32:09] have all these test functions to be able to characterize it it needs to be
[01:32:10] to characterize it it needs to be flexible because you you go to different
[01:32:12] flexible because you you go to different you transition from different stages
[01:32:15] you transition from different stages product life cycle manufacturing cycle
[01:32:17] product life cycle manufacturing cycle you know it needs to be integrated into
[01:32:19] you know it needs to be integrated into the the probing and assembly equipment
[01:32:21] the the probing and assembly equipment very important you know scalable to high
[01:32:24] very important you know scalable to high Channel count and parallel testing you
[01:32:26] Channel count and parallel testing you know high density because you know you
[01:32:28] know high density because you know you need to do a lot of testing you need to
[01:32:30] need to do a lot of testing you need to integrate it into these into these
[01:32:31] integrate it into these into these machines so that you don't want to take
[01:32:33] machines so that you don't want to take up too much space that's that's pricey
[01:32:35] up too much space that's that's pricey and of course optimize test slow to
[01:32:38] and of course optimize test slow to really test your device as quickly as
[01:32:39] really test your device as quickly as possible so what are the ask for the
[01:32:42] possible so what are the ask for the ecosystems why do we need to cooperate
[01:32:44] ecosystems why do we need to cooperate and this has already been mentioned um
[01:32:46] and this has already been mentioned um by some of the other partners is you
[01:32:48] by some of the other partners is you know what what kind of standardization
[01:32:50] know what what kind of standardization do we need to have like Optical signal
[01:32:52] do we need to have like Optical signal access you know it's it's not trivial
[01:32:55] access you know it's it's not trivial it's not just like Electronics it's it's
[01:32:57] it's not just like Electronics it's it's a little bit more complicated are we
[01:32:59] a little bit more complicated are we going to standardize on that you know
[01:33:02] going to standardize on that you know the the measurements and the test plan
[01:33:04] the the measurements and the test plan it seemed that we're kind of converging
[01:33:05] it seemed that we're kind of converging on something um and which is really
[01:33:08] on something um and which is really powerful so that we know exactly what
[01:33:10] powerful so that we know exactly what needs to be tested and when and then you
[01:33:13] needs to be tested and when and then you know maybe some thoughts like do you
[01:33:14] know maybe some thoughts like do you want to have a standardized industry
[01:33:17] want to have a standardized industry Hardware framework that has certain
[01:33:20] Hardware framework that has certain mechanical uh you know Dimensions just
[01:33:22] mechanical uh you know Dimensions just mod kind of a modular approach something
[01:33:24] mod kind of a modular approach something like pxi where you have you know
[01:33:26] like pxi where you have you know Hardware triggers so you can really
[01:33:28] Hardware triggers so you can really optimize your test flow and also maybe
[01:33:30] optimize your test flow and also maybe kind of a software framework where you
[01:33:33] kind of a software framework where you have you know kind of an API structure
[01:33:35] have you know kind of an API structure but maybe kind of a modern version of of
[01:33:38] but maybe kind of a modern version of of Skippy like G RPC to make the test flow
[01:33:42] Skippy like G RPC to make the test flow you know as as Optimum as possible so
[01:33:45] you know as as Optimum as possible so this kind of concludes my my
[01:33:47] this kind of concludes my my presentation on a testing perspective on
[01:33:50] presentation on a testing perspective on scaling inter integrated photonics
[01:33:52] scaling inter integrated photonics Technologies thank you very much case I
[01:33:54] Technologies thank you very much case I can tell you AI is changing the game
[01:33:57] can tell you AI is changing the game right now ai is changing the game
[01:33:58] right now ai is changing the game actually last year at ofc we had a Labs
[01:34:01] actually last year at ofc we had a Labs talking to about 4 terab per second wdm
[01:34:03] talking to about 4 terab per second wdm needed for AI at the PEC Summit last
[01:34:07] needed for AI at the PEC Summit last last October we have a video telling us
[01:34:09] last October we have a video telling us about multi terabit per second the high
[01:34:11] about multi terabit per second the high speeds are are crazy how is quantify
[01:34:16] speeds are are crazy how is quantify strategizing Towers offering the testing
[01:34:18] strategizing Towers offering the testing solutions that the AI is going to need
[01:34:20] solutions that the AI is going to need today
[01:34:22] today yeah it's a great question I mean right
[01:34:24] yeah it's a great question I mean right now what we're seeing is that the um you
[01:34:27] now what we're seeing is that the um you know there's still a lot of Technology
[01:34:29] know there's still a lot of Technology at at 100g per Channel U that is going
[01:34:32] at at 100g per Channel U that is going to be scaled in in the next in the next
[01:34:34] to be scaled in in the next in the next couple of years and of course then the
[01:34:36] couple of years and of course then the next wave is going to be 200g per
[01:34:38] next wave is going to be 200g per Channel and you know we we are actually
[01:34:41] Channel and you know we we are actually developing solutions to to be able to
[01:34:43] developing solutions to to be able to align very well with with with that kind
[01:34:46] align very well with with with that kind of Technology road map so very important
[01:34:48] of Technology road map so very important um you know but again 100g is is is
[01:34:50] um you know but again 100g is is is probably going to be the first phase
[01:34:52] probably going to be the first phase will will scaling in in the next couple
[01:34:54] will will scaling in in the next couple of years and then 200g maybe a few years
[01:34:58] of years and then 200g maybe a few years later I would like to to now continue
[01:35:01] later I would like to to now continue with the gr to link all this ecosystem
[01:35:04] with the gr to link all this ecosystem to one company who really has put their
[01:35:07] to one company who really has put their money where the mouth is we go to
[01:35:09] money where the mouth is we go to broadcom you know in September there was
[01:35:10] broadcom you know in September there was this big press release with broadcom go
[01:35:13] this big press release with broadcom go talking to Google Cloud for was
[01:35:15] talking to Google Cloud for was generating an ecosystem for generative
[01:35:17] generating an ecosystem for generative Ai and they need a physical layer they
[01:35:19] Ai and they need a physical layer they need to provide they need to technology
[01:35:22] need to provide they need to technology that will enable AI in the future for
[01:35:24] that will enable AI in the future for that we have Manish mea with us from
[01:35:26] that we have Manish mea with us from broadcom to close the meeting the way
[01:35:28] broadcom to close the meeting the way that it is Manish before I give you the
[01:35:30] that it is Manish before I give you the floor one thing for everybody in the
[01:35:32] floor one thing for everybody in the room if you got questions for Manish or
[01:35:34] room if you got questions for Manish or for any of one of the other speakers now
[01:35:37] for any of one of the other speakers now it is a time to put them in the Q&A
[01:35:40] it is a time to put them in the Q&A section press the Q&A button and write
[01:35:42] section press the Q&A button and write your question there after manage
[01:35:44] your question there after manage presentation we'll have our q& which is
[01:35:46] presentation we'll have our q& which is of course for me the most exciting part
[01:35:47] of course for me the most exciting part of the meeting but now we go to broadcom
[01:35:49] of the meeting but now we go to broadcom manage the floor at the attention of
[01:35:51] manage the floor at the attention of every one is yours I thank you Jose and
[01:35:54] every one is yours I thank you Jose and can you hear me clearly and see the
[01:35:55] can you hear me clearly and see the slides I'm clear okay great so uh first
[01:35:59] slides I'm clear okay great so uh first of all want to thank APC and want to
[01:36:01] of all want to thank APC and want to thank Jose for for putting on this
[01:36:03] thank Jose for for putting on this session uh and happy New Year to
[01:36:06] session uh and happy New Year to everyone um you know I think at broadcom
[01:36:08] everyone um you know I think at broadcom we've been quite public about our CPO
[01:36:11] we've been quite public about our CPO Investments for the last several years
[01:36:13] Investments for the last several years I'll share some of our technical
[01:36:15] I'll share some of our technical progress and uh share some thoughts on
[01:36:17] progress and uh share some thoughts on ecosystem development uh towards the end
[01:36:20] ecosystem development uh towards the end of the uh presentation so this is a
[01:36:23] of the uh presentation so this is a slide that I think many of you have have
[01:36:25] slide that I think many of you have have probably seen before or at least a
[01:36:27] probably seen before or at least a similar version and at broadcom we
[01:36:30] similar version and at broadcom we focused our silic and photonics
[01:36:31] focused our silic and photonics investment on co-packaged Optics uh
[01:36:34] investment on co-packaged Optics uh initially targeting to pair targeting to
[01:36:37] initially targeting to pair targeting to pair our Optical engines with broadcom
[01:36:40] pair our Optical engines with broadcom core silicon for networking applications
[01:36:43] core silicon for networking applications and the target value proposition is
[01:36:45] and the target value proposition is wellknown lower cost and lower power
[01:36:47] wellknown lower cost and lower power compared to traditional Optical
[01:36:49] compared to traditional Optical interconnects and ultimately we believe
[01:36:52] interconnects and ultimately we believe delivering silicon level reliability uh
[01:36:55] delivering silicon level reliability uh the graphic on the right shows our
[01:36:58] the graphic on the right shows our estimates for the for the uh Power
[01:37:00] estimates for the for the uh Power benefit um against other Optical
[01:37:03] benefit um against other Optical interconnect Solutions today for
[01:37:04] interconnect Solutions today for networking where we expect the
[01:37:06] networking where we expect the co-packaged Optics to deliver a 60 to
[01:37:09] co-packaged Optics to deliver a 60 to 70% reduction uh compared to traditional
[01:37:14] 70% reduction uh compared to traditional Optics and and this um you know slide
[01:37:17] Optics and and this um you know slide just visualizes the migration uh of
[01:37:20] just visualizes the migration uh of course I'm sure many of you are aware of
[01:37:22] course I'm sure many of you are aware of traditional Optics to co-packaged Optics
[01:37:25] traditional Optics to co-packaged Optics uh the left side shows an explosion of
[01:37:26] uh the left side shows an explosion of one of our standard uh single mode 4
[01:37:29] one of our standard uh single mode 4 channel modules and while this Paradigm
[01:37:32] channel modules and while this Paradigm has deployed widely in the industry for
[01:37:33] has deployed widely in the industry for more than 20 years it is discreet it's
[01:37:36] more than 20 years it is discreet it's manual labor intensive and the industry
[01:37:39] manual labor intensive and the industry has made uh pretty great strides in the
[01:37:41] has made uh pretty great strides in the last 7 eight years uh progressing to
[01:37:43] last 7 eight years uh progressing to what you see in the middle uh ramping
[01:37:46] what you see in the middle uh ramping silic and photonics however it has been
[01:37:48] silic and photonics however it has been predominantly with um uh low density
[01:37:51] predominantly with um uh low density Optical engines used in uh transceivers
[01:37:54] Optical engines used in uh transceivers you know either four Channel and more
[01:37:57] you know either four Channel and more recently eight channel devices using
[01:37:59] recently eight channel devices using silicon photonics so our goal is to now
[01:38:03] silicon photonics so our goal is to now take that one step further uh increase
[01:38:05] take that one step further uh increase the integration density by an order of
[01:38:08] the integration density by an order of magnitude and then use Advanced
[01:38:10] magnitude and then use Advanced semiconductor packaging uh to co-
[01:38:12] semiconductor packaging uh to co- package the Optics on a common substrate
[01:38:15] package the Optics on a common substrate with core networking
[01:38:17] with core networking as6 so of course there are you know many
[01:38:20] as6 so of course there are you know many ways to the problems and and there are
[01:38:22] ways to the problems and and there are many solutions that different companies
[01:38:24] many solutions that different companies and Industry can and should try um you
[01:38:28] and Industry can and should try um you know innovation in the space does need
[01:38:30] know innovation in the space does need to run in parallel by various parties um
[01:38:34] to run in parallel by various parties um some of the areas that we focused uh on
[01:38:36] some of the areas that we focused uh on our first CPO program that we called
[01:38:39] our first CPO program that we called Humbolt uh was to develop Optical
[01:38:41] Humbolt uh was to develop Optical engines with 32 Channel photonic circuit
[01:38:43] engines with 32 Channel photonic circuit die bonded to 32 Channel electrical
[01:38:47] die bonded to 32 Channel electrical circuit die the electrical circuit die
[01:38:49] circuit die the electrical circuit die have our modulator drivers for the TX
[01:38:52] have our modulator drivers for the TX path um and TI for the RX path we also
[01:38:56] path um and TI for the RX path we also developed a detachable optical fiber
[01:38:58] developed a detachable optical fiber connection uh at the optical engine
[01:39:00] connection uh at the optical engine interface um in order to facilitate a
[01:39:03] interface um in order to facilitate a soldered Optical attach uh to the
[01:39:06] soldered Optical attach uh to the substrate and we are soldered attach of
[01:39:09] substrate and we are soldered attach of the optical engine to the substrate and
[01:39:11] the optical engine to the substrate and we also worked pretty early on with odm
[01:39:14] we also worked pretty early on with odm Partners to test uh system integration
[01:39:17] Partners to test uh system integration in a chassis manufacturing environment
[01:39:22] in a chassis manufacturing environment um so here this uh shows you know an
[01:39:27] um so here this uh shows you know an example of the results of one of our
[01:39:29] example of the results of one of our Solutions uh the left picture shows a a
[01:39:32] Solutions uh the left picture shows a a standard switch uh which is you know
[01:39:34] standard switch uh which is you know populated by a large number say 32 to
[01:39:37] populated by a large number say 32 to 128 Optical transceivers the high-speed
[01:39:40] 128 Optical transceivers the high-speed signals propagate from the core Asic um
[01:39:43] signals propagate from the core Asic um through the host PCB in the box to the
[01:39:45] through the host PCB in the box to the optical transceivers whereas the right
[01:39:48] optical transceivers whereas the right picture shows our Humbolt uh CP
[01:39:51] picture shows our Humbolt uh CP uh svk system the validation vehicle web
[01:39:54] uh svk system the validation vehicle web built and this system was a half CPO and
[01:39:58] built and this system was a half CPO and half pluggable but you can see the
[01:40:00] half pluggable but you can see the implementation of the CPO section um
[01:40:03] implementation of the CPO section um where the highspeed signals propagate
[01:40:05] where the highspeed signals propagate from the core as6 to the face plate over
[01:40:09] from the core as6 to the face plate over uh
[01:40:10] uh fiber and then this um you know is an
[01:40:13] fiber and then this um you know is an example of what that half CPO half
[01:40:15] example of what that half CPO half plugable implementation looks like in an
[01:40:17] plugable implementation looks like in an odm box on the front panel uh the top
[01:40:21] odm box on the front panel uh the top section shows the CPO with fiber IO
[01:40:24] section shows the CPO with fiber IO ports and remote laser module uh pluga
[01:40:27] ports and remote laser module uh pluga bles on the edges and the bottom shows
[01:40:30] bles on the edges and the bottom shows uh the standard pluggable section and
[01:40:32] uh the standard pluggable section and and you can kind of visually see um how
[01:40:35] and you can kind of visually see um how moving to CPO from standard pluggable
[01:40:38] moving to CPO from standard pluggable gives pretty significant face plate
[01:40:39] gives pretty significant face plate Savings in this case it's about 30%
[01:40:42] Savings in this case it's about 30% additional face plate area for
[01:40:44] additional face plate area for ventilation um that does offer some
[01:40:46] ventilation um that does offer some pretty interesting system design
[01:40:49] pretty interesting system design opportunities and um and you know wanted
[01:40:53] opportunities and um and you know wanted to share won't go into too much detail
[01:40:55] to share won't go into too much detail here but wanted to share uh some results
[01:40:58] here but wanted to share uh some results from our early production builds uh
[01:41:01] from our early production builds uh these are not one-off builds you know
[01:41:02] these are not one-off builds you know these are multiple Lots uh multiple
[01:41:04] these are multiple Lots uh multiple devices and a lot that uh we have run
[01:41:07] devices and a lot that uh we have run through uh manufacturing
[01:41:10] through uh manufacturing processes uh and you know I think we've
[01:41:13] processes uh and you know I think we've we've been able to get to consistency on
[01:41:16] we've been able to get to consistency on optical engine performance from run to
[01:41:18] optical engine performance from run to run and if you look at the bottom right
[01:41:20] run and if you look at the bottom right plot
[01:41:21] plot um you know this was our first
[01:41:22] um you know this was our first generation silicon germanium based
[01:41:25] generation silicon germanium based device where uh we did uh have wor case
[01:41:29] device where uh we did uh have wor case power that was under eight Watts um in
[01:41:31] power that was under eight Watts um in full system op operation uh eight watts
[01:41:34] full system op operation uh eight watts per 800 gig for the optical interconnect
[01:41:38] per 800 gig for the optical interconnect uh we do expect that to come down
[01:41:40] uh we do expect that to come down further on uh the most recent devices
[01:41:44] further on uh the most recent devices that we're building which are based on
[01:41:45] that we're building which are based on seos
[01:41:47] seos implementations so um now I do think you
[01:41:51] implementations so um now I do think you know aside from all of the Upstream
[01:41:53] know aside from all of the Upstream challenges that certainly exist in in uh
[01:41:56] challenges that certainly exist in in uh developing and deploying highly
[01:41:59] developing and deploying highly integrated silicon photonics there has
[01:42:02] integrated silicon photonics there has also been a lot of fear in the industry
[01:42:05] also been a lot of fear in the industry on what it would be like to assemble
[01:42:07] on what it would be like to assemble bring up and deploy a CPO system in a
[01:42:11] bring up and deploy a CPO system in a network since that architecture is so
[01:42:14] network since that architecture is so new for Optics and honestly it was it
[01:42:16] new for Optics and honestly it was it was pretty scary for us as well early on
[01:42:19] was pretty scary for us as well early on it took us three months
[01:42:21] it took us three months um to bring up the CPO in a system the
[01:42:25] um to bring up the CPO in a system the first time that we had a successful
[01:42:28] first time that we had a successful installation um within a year we had
[01:42:30] installation um within a year we had been able to reduce that bring up time
[01:42:33] been able to reduce that bring up time to about two weeks which was a
[01:42:36] to about two weeks which was a significant Improvement but still
[01:42:38] significant Improvement but still nowhere near acceptable for scale
[01:42:41] nowhere near acceptable for scale deployment um and then it took another
[01:42:43] deployment um and then it took another six months after that to where we could
[01:42:45] six months after that to where we could bring up a system within a day um but of
[01:42:47] bring up a system within a day um but of course now that we've gone through that
[01:42:50] course now that we've gone through that um experience and gone through that
[01:42:51] um experience and gone through that learning curve um we you know feel like
[01:42:55] learning curve um we you know feel like that that's something that is is no
[01:42:57] that that's something that is is no longer um will be able to more quickly
[01:43:01] longer um will be able to more quickly get to those shorter time
[01:43:04] get to those shorter time Cycles so you know we continue uh to
[01:43:07] Cycles so you know we continue uh to make progress on our programs you know
[01:43:09] make progress on our programs you know we've got a good base of customers both
[01:43:11] we've got a good base of customers both system integrators and end users uh
[01:43:14] system integrators and end users uh planning deployments but there are some
[01:43:16] planning deployments but there are some areas uh of Industry collaboration that
[01:43:19] areas uh of Industry collaboration that can accelerate time to high volume from
[01:43:21] can accelerate time to high volume from you know our perspective I'll share some
[01:43:23] you know our perspective I'll share some of those thoughts in the next few slides
[01:43:26] of those thoughts in the next few slides um you know these these are some images
[01:43:28] um you know these these are some images that illustrate the optical path of you
[01:43:30] that illustrate the optical path of you know the types of systems that we're
[01:43:32] know the types of systems that we're developing and you know we start with an
[01:43:34] developing and you know we start with an optical engine you can see a cartoon on
[01:43:36] optical engine you can see a cartoon on top um that has our pick and our EIC um
[01:43:41] top um that has our pick and our EIC um you know we we connect them using fan
[01:43:43] you know we we connect them using fan out wa for level processing or fowlp
[01:43:46] out wa for level processing or fowlp which is um you know a high highly
[01:43:49] which is um you know a high highly integrated Advanced Packaging technique
[01:43:51] integrated Advanced Packaging technique um
[01:43:52] um deployed um which is critical for for
[01:43:56] deployed um which is critical for for the type of density of if integration
[01:43:58] the type of density of if integration and bumps that we need or or connections
[01:44:00] and bumps that we need or or connections that we need between the two chips and
[01:44:02] that we need between the two chips and then we Bond these engines to a common
[01:44:04] then we Bond these engines to a common substrate with our core ASD uh within
[01:44:07] substrate with our core ASD uh within the Box the optical engines route to
[01:44:09] the Box the optical engines route to both the front panel IO ports and remote
[01:44:12] both the front panel IO ports and remote laser module uh Optical connectors um
[01:44:15] laser module uh Optical connectors um now we've been targeting networking
[01:44:17] now we've been targeting networking deployments uh but one place that you
[01:44:19] deployments uh but one place that you know I think would be good to begin
[01:44:21] know I think would be good to begin revisiting commonality is what's
[01:44:23] revisiting commonality is what's actually progressing down the fiber we
[01:44:25] actually progressing down the fiber we are progressing uh in the industry right
[01:44:28] are progressing uh in the industry right now in a direction where compute
[01:44:30] now in a direction where compute applications and networking applications
[01:44:33] applications and networking applications will require very different wavelength
[01:44:35] will require very different wavelength plans and uh and because of that uh
[01:44:38] plans and uh and because of that uh integration schemes and I I do think
[01:44:41] integration schemes and I I do think there's you know important discussion
[01:44:43] there's you know important discussion needed on whether the application
[01:44:45] needed on whether the application scenarios that are driving this
[01:44:48] scenarios that are driving this variation um are strong enough to
[01:44:50] variation um are strong enough to support
[01:44:52] support fragmentation um because you know again
[01:44:54] fragmentation um because you know again there is there's a broadly um uh
[01:44:58] there is there's a broadly um uh deployed base of uh kind of parallel
[01:45:01] deployed base of uh kind of parallel single mode uh four channel uh four
[01:45:05] single mode uh four channel uh four wavelength solutions that for um
[01:45:09] wavelength solutions that for um networking or even a lot of the emerging
[01:45:10] networking or even a lot of the emerging AI applications which are still highly
[01:45:13] AI applications which are still highly networking oriented um uh can continue
[01:45:16] networking oriented um uh can continue to progress down this path you know
[01:45:18] to progress down this path you know moving from 100 Gig to 200 g per Lane uh
[01:45:23] moving from 100 Gig to 200 g per Lane uh speeds um we've also described a fair
[01:45:26] speeds um we've also described a fair amount in the past about our remote
[01:45:28] amount in the past about our remote laser module design now frankly we we
[01:45:31] laser module design now frankly we we were already progressing quickly when
[01:45:33] were already progressing quickly when the elsf standard was emerging and our
[01:45:35] the elsf standard was emerging and our focus at the time was to deliver remote
[01:45:38] focus at the time was to deliver remote laser modules with uncooled 20 dbm
[01:45:41] laser modules with uncooled 20 dbm lasers um but the rlm is an important
[01:45:43] lasers um but the rlm is an important area of development in in the industry
[01:45:46] area of development in in the industry um you know we'd like to see continued
[01:45:48] um you know we'd like to see continued progress uh increasing laser power and
[01:45:51] progress uh increasing laser power and enabling either longer link budgets or
[01:45:53] enabling either longer link budgets or higher split ratios you know therefore
[01:45:55] higher split ratios you know therefore reducing the amount of fiber within
[01:45:58] reducing the amount of fiber within that's needed in a box um we are open to
[01:46:01] that's needed in a box um we are open to integrating our CPO Solutions with
[01:46:04] integrating our CPO Solutions with various rlm designs and form factors it
[01:46:07] various rlm designs and form factors it does still feel early to decide on a
[01:46:09] does still feel early to decide on a single solution um we don't know and I
[01:46:12] single solution um we don't know and I don't think the industry knows yet
[01:46:14] don't think the industry knows yet whether the right optimization vectors
[01:46:16] whether the right optimization vectors are say cooled versus uncooled or laser
[01:46:19] are say cooled versus uncooled or laser diode power versus quantity but um I
[01:46:22] diode power versus quantity but um I think we're pretty confident and do
[01:46:24] think we're pretty confident and do believe that this is definitely a part
[01:46:26] believe that this is definitely a part of a a CPO solution that will uh
[01:46:29] of a a CPO solution that will uh disaggregate and will uh get you know
[01:46:32] disaggregate and will uh get you know will benefit
[01:46:33] will benefit from um from interoperability and and
[01:46:37] from um from interoperability and and more uh industry level of
[01:46:40] more uh industry level of evolation um at the CPO or packaging
[01:46:43] evolation um at the CPO or packaging level you know we do hope to see more
[01:46:44] level you know we do hope to see more emphasis by Founders noats keeping
[01:46:47] emphasis by Founders noats keeping photonics in mind uh during roadmap
[01:46:49] photonics in mind uh during roadmap development um to support both silicon
[01:46:52] development um to support both silicon plutonics and CPR requirements some
[01:46:54] plutonics and CPR requirements some examples of that you know double side
[01:46:57] examples of that you know double side attach when you're using say uh you know
[01:47:00] attach when you're using say uh you know pick EIC bonding rather than monolithic
[01:47:02] pick EIC bonding rather than monolithic Solutions so that you can take advantage
[01:47:04] Solutions so that you can take advantage of you know low speed very very Advanced
[01:47:08] of you know low speed very very Advanced uh Coss nodes five and 7even animet for
[01:47:10] uh Coss nodes five and 7even animet for low power Electronics coupled with
[01:47:13] low power Electronics coupled with optimized you know photonic circuits um
[01:47:16] optimized you know photonic circuits um you need to attach you know on one side
[01:47:18] you need to attach you know on one side to a substrate or or some type package
[01:47:20] to a substrate or or some type package on the other side to the other device
[01:47:23] on the other side to the other device and um that double side attach is is
[01:47:25] and um that double side attach is is often unique relative to uh standard
[01:47:28] often unique relative to uh standard electronic Solutions and you know we've
[01:47:31] electronic Solutions and you know we've also observed cleanliness requirements
[01:47:34] also observed cleanliness requirements uh needing some more I guess investment
[01:47:36] uh needing some more I guess investment and discussion um in the industry uh as
[01:47:40] and discussion um in the industry uh as a lot of the cleanliness requirements
[01:47:42] a lot of the cleanliness requirements that are needed for Optics are were not
[01:47:44] that are needed for Optics are were not necessary for
[01:47:47] necessary for electronics and then finally um you know
[01:47:50] electronics and then finally um you know system implementations using CPO are
[01:47:53] system implementations using CPO are very new and you know the industry has
[01:47:56] very new and you know the industry has become comfortable using gr 468 to
[01:47:59] become comfortable using gr 468 to specify Optical transceiver reliability
[01:48:02] specify Optical transceiver reliability uh the industry obviously has agreement
[01:48:03] uh the industry obviously has agreement on system level strife and real but uh
[01:48:06] on system level strife and real but uh embedding Optics in a chassis does
[01:48:08] embedding Optics in a chassis does require new stress protocols at the
[01:48:10] require new stress protocols at the system level uh we're happy within
[01:48:12] system level uh we're happy within broadcom to share our learnings is we
[01:48:14] broadcom to share our learnings is we also become more mature uh and APC would
[01:48:17] also become more mature uh and APC would be a great place to do this um but but
[01:48:19] be a great place to do this um but but this is again another area that I think
[01:48:22] this is again another area that I think um if we look further Downstream will
[01:48:25] um if we look further Downstream will require some uh more industry discussion
[01:48:28] require some uh more industry discussion and in
[01:48:29] and in collaboration um you know I think our
[01:48:31] collaboration um you know I think our work I'd like to believe our work thus
[01:48:33] work I'd like to believe our work thus far has shown that there are no more
[01:48:34] far has shown that there are no more remaining technical showstoppers uh to
[01:48:37] remaining technical showstoppers uh to deploying true co-packaging and I think
[01:48:40] deploying true co-packaging and I think that that is a meaningful step um
[01:48:42] that that is a meaningful step um furthermore the appliance ecosystem or
[01:48:44] furthermore the appliance ecosystem or some of the appliance ecosystem has
[01:48:46] some of the appliance ecosystem has embraced CPO and they are getting a leg
[01:48:48] embraced CPO and they are getting a leg up on development and deployment uh but
[01:48:51] up on development and deployment uh but most importantly bring up is becoming
[01:48:53] most importantly bring up is becoming more familiar to a wider range of
[01:48:55] more familiar to a wider range of developers uh you know next key steps
[01:48:57] developers uh you know next key steps are are reliability real yield
[01:49:02] are are reliability real yield repeatability um but and of course with
[01:49:04] repeatability um but and of course with that there's going to be more ecosystem
[01:49:05] that there's going to be more ecosystem opportunity for for companies to come in
[01:49:08] opportunity for for companies to come in and and develop uh critical new
[01:49:09] and and develop uh critical new solutions that are required but um I
[01:49:13] solutions that are required but um I think uh looking forward to where where
[01:49:16] think uh looking forward to where where forums like APC can take the the next
[01:49:18] forums like APC can take the the next phases of Industry collaboration
[01:49:21] phases of Industry collaboration so thank you thank you very much Manish
[01:49:24] so thank you thank you very much Manish thank you for a great presentation but
[01:49:25] thank you for a great presentation but also thank you from the very beginning
[01:49:27] also thank you from the very beginning Believing on C- package Optics I think
[01:49:30] Believing on C- package Optics I think everybody knows this about me I'm a
[01:49:32] everybody knows this about me I'm a really huge fan of CPO I believe there
[01:49:34] really huge fan of CPO I believe there were many and there are still many
[01:49:36] were many and there are still many challenges that we need to solve but
[01:49:38] challenges that we need to solve but this in my mind the most beautiful way
[01:49:40] this in my mind the most beautiful way of solving it the lack of the the the
[01:49:42] of solving it the lack of the the the increased density of photonics at the
[01:49:45] increased density of photonics at the data center level so thank you very much
[01:49:46] data center level so thank you very much Manish let's start the Q&A so let me ask
[01:49:49] Manish let's start the Q&A so let me ask all our all other speakers to switch on
[01:49:51] all our all other speakers to switch on the camera and look their best for the
[01:49:53] the camera and look their best for the next 15 minutes the first question is
[01:49:56] next 15 minutes the first question is from johi from Photon Del and goes to
[01:49:58] from johi from Photon Del and goes to Jeffrey Maki is APC restricted to IND to
[01:50:03] Jeffrey Maki is APC restricted to IND to Silicon photonics or are other materials
[01:50:05] Silicon photonics or are other materials like for example indium phosphate part
[01:50:07] like for example indium phosphate part of your
[01:50:08] of your ecosystem we are not restricted uh we
[01:50:12] ecosystem we are not restricted uh we presented a a mission that we're working
[01:50:14] presented a a mission that we're working on to make something happen in the
[01:50:16] on to make something happen in the industry if there's interest for Indian
[01:50:18] industry if there's interest for Indian fosi monolithic integration uh project
[01:50:22] fosi monolithic integration uh project of some kind certainly we're open to it
[01:50:24] of some kind certainly we're open to it perhaps uh within our efforts we'll find
[01:50:27] perhaps uh within our efforts we'll find that some standards could be equally
[01:50:30] that some standards could be equally phrased for silicon botonics and indium
[01:50:32] phrased for silicon botonics and indium phosphide and certainly as we're here in
[01:50:34] phosphide and certainly as we're here in our presentation today we're still using
[01:50:35] our presentation today we're still using indium F we need
[01:50:37] indium F we need lasers the next question goes to Global
[01:50:41] lasers the next question goes to Global fundies the GF photonic pdk Does it
[01:50:45] fundies the GF photonic pdk Does it include heaters are they under
[01:50:47] include heaters are they under development when are we going to have
[01:50:48] development when are we going to have heaters in your silicon Bic platform
[01:50:51] heaters in your silicon Bic platform they're already available in the pdk and
[01:50:53] they're already available in the pdk and they been I think there was a question
[01:50:54] they been I think there was a question also about the reliability they've been
[01:50:57] also about the reliability they've been pre-characterized as I'd mentioned
[01:50:58] pre-characterized as I'd mentioned before we're trying to do this work
[01:51:00] before we're trying to do this work ahead of people using using these
[01:51:03] ahead of people using using these components so they're already in the pdk
[01:51:04] components so they're already in the pdk including the undercut how how fast can
[01:51:08] including the undercut how how fast can you tune with heaters is it
[01:51:10] you tune with heaters is it kilohertz mehz you would have to refer
[01:51:12] kilohertz mehz you would have to refer to it in the
[01:51:14] to it in the pdk of course and the next question goes
[01:51:16] pdk of course and the next question goes to the oset in the room we go to das and
[01:51:18] to the oset in the room we go to das and Cassandra citronic
[01:51:20] Cassandra citronic can you talk a little bit more about the
[01:51:22] can you talk a little bit more about the wafer level test where the optical
[01:51:24] wafer level test where the optical signal the laser is piped in
[01:51:28] signal the laser is piped in externally so on that question what I
[01:51:30] externally so on that question what I would say is the main challenges of that
[01:51:32] would say is the main challenges of that are similar to that of assembly in that
[01:51:34] are similar to that of assembly in that there are just so many different designs
[01:51:37] there are just so many different designs everything's very unique so having all
[01:51:39] everything's very unique so having all of the tooling uh is less possible to
[01:51:44] of the tooling uh is less possible to have all at once just given that you
[01:51:46] have all at once just given that you know grading couplers Edge couplers and
[01:51:48] know grading couplers Edge couplers and any type of electrical Pro probing just
[01:51:50] any type of electrical Pro probing just everything's very very different at all
[01:51:53] everything's very very different at all times the next question I go back to
[01:51:55] times the next question I go back to Global fundra I go back to vas say
[01:51:58] Global fundra I go back to vas say twofold question ER can we use your pdk
[01:52:02] twofold question ER can we use your pdk with different software vendors not only
[01:52:05] with different software vendors not only the small one but the big ones and of
[01:52:08] the small one but the big ones and of course can we also H use your pdk to
[01:52:14] course can we also H use your pdk to um sorry are there any schedule for mpw
[01:52:18] um sorry are there any schedule for mpw runs today and what is it yeah so I
[01:52:22] runs today and what is it yeah so I think the the mpw runs I think there is
[01:52:24] think the the mpw runs I think there is uh the the schedules are on the GF
[01:52:27] uh the the schedules are on the GF external website and and uh if you can't
[01:52:30] external website and and uh if you can't find it just send me an email I can put
[01:52:32] find it just send me an email I can put you in touch with our salespeople who
[01:52:34] you in touch with our salespeople who can give you the mpw
[01:52:37] can give you the mpw schedule all right let's go to the next
[01:52:40] schedule all right let's go to the next the the next question and the next
[01:52:41] the the next question and the next question goes to javil let's see how we
[01:52:44] question goes to javil let's see how we address this one because I think it's a
[01:52:46] address this one because I think it's a very good one no you mentioned the
[01:52:48] very good one no you mentioned the chicken situation between capex needed
[01:52:51] chicken situation between capex needed to support hbm and the need for customer
[01:52:53] to support hbm and the need for customer demand do you see interest from
[01:52:55] demand do you see interest from customers to co-invest and if so in
[01:52:59] customers to co-invest and if so in which applications and if not is it
[01:53:02] which applications and if not is it because current Tech is considered just
[01:53:04] because current Tech is considered just good
[01:53:06] enough uh I will give a very limited
[01:53:09] enough uh I will give a very limited answer on this
[01:53:11] answer on this um when
[01:53:13] um when clients have a very bespoke solution
[01:53:16] clients have a very bespoke solution requires custom tooling that needs to be
[01:53:19] requires custom tooling that needs to be is developed just for their product then
[01:53:22] is developed just for their product then yeah there's a either they'll consign it
[01:53:25] yeah there's a either they'll consign it or there'll be some
[01:53:26] or there'll be some agreement I think in order to get to the
[01:53:30] agreement I think in order to get to the cost targets that we
[01:53:33] cost targets that we want we will need some
[01:53:36] want we will need some standardization and then it cannot be
[01:53:38] standardization and then it cannot be super visis spoke it needs to be with
[01:53:40] super visis spoke it needs to be with some
[01:53:41] some commonality uh to
[01:53:44] commonality uh to others the other part of it that I will
[01:53:46] others the other part of it that I will kind of address is that at the end that
[01:53:50] kind of address is that at the end that the fiber attach at the latter parts of
[01:53:53] the fiber attach at the latter parts of assembly the
[01:53:55] assembly the minimum uh minimum capacity per tool is
[01:54:01] minimum uh minimum capacity per tool is smaller side so you can have an easier
[01:54:03] smaller side so you can have an easier ramp up that you buy a certain size of
[01:54:05] ramp up that you buy a certain size of Fleet of tools and then you gradually
[01:54:07] Fleet of tools and then you gradually increase it as you creep up uh closer to
[01:54:11] increase it as you creep up uh closer to when you look at these Advanced
[01:54:12] when you look at these Advanced packaging schemes that have to do you
[01:54:14] packaging schemes that have to do you have to be able to handle Wafers tooling
[01:54:17] have to be able to handle Wafers tooling becomes much more expensive and the
[01:54:20] becomes much more expensive and the and the throughput of every one tool is
[01:54:23] and the throughput of every one tool is kind of geared towards the seos world
[01:54:25] kind of geared towards the seos world and capacities there so it's harder to
[01:54:29] and capacities there so it's harder to break it up and do a gradual
[01:54:31] break it up and do a gradual investment
[01:54:34] investment um yeah I don't think I could comment
[01:54:37] um yeah I don't think I could comment exactly on specific clients and yeah did
[01:54:40] exactly on specific clients and yeah did Ash really wants to comment on this I
[01:54:43] Ash really wants to comment on this I believe yes yes so you know in my past
[01:54:46] believe yes yes so you know in my past life I used to work for AC which is
[01:54:49] life I used to work for AC which is number one in assembly and test we have
[01:54:52] number one in assembly and test we have ramp up many devices at weer level as
[01:54:54] ramp up many devices at weer level as well as packaging so let me address
[01:54:57] well as packaging so let me address between the traditional Electronics IC
[01:55:00] between the traditional Electronics IC and Optical packaging uh so in uh
[01:55:03] and Optical packaging uh so in uh traditional IC you the bump size as Vias
[01:55:06] traditional IC you the bump size as Vias may say is pretty much perfect pretty
[01:55:08] may say is pretty much perfect pretty much fixed normally it's uh 130 Micron
[01:55:11] much fixed normally it's uh 130 Micron page sometime it's 150 for copper pillar
[01:55:14] page sometime it's 150 for copper pillar is normally 80 Micron page for the ball
[01:55:17] is normally 80 Micron page for the ball size it's like2 3.4 is pretty fixed
[01:55:21] size it's like2 3.4 is pretty fixed wieter is pretty fixed so all these in
[01:55:23] wieter is pretty fixed so all these in electronics IC packaging is already
[01:55:26] electronics IC packaging is already fixed so whether Nvidia or broadcom or
[01:55:29] fixed so whether Nvidia or broadcom or Qualcomm Apple anybody designing
[01:55:31] Qualcomm Apple anybody designing Electronics IC the package size is
[01:55:33] Electronics IC the package size is perfect it's fixed they know what is a
[01:55:35] perfect it's fixed they know what is a design Rule and they can ramp up on the
[01:55:38] design Rule and they can ramp up on the photonex IC we have worked with multiple
[01:55:40] photonex IC we have worked with multiple customers none of even customers among
[01:55:43] customers none of even customers among themselves when they're going from 100g
[01:55:45] themselves when they're going from 100g transfer module to 200g transfer module
[01:55:48] transfer module to 200g transfer module the dice size is different the the uh
[01:55:50] the dice size is different the the uh the number of fa is different the size
[01:55:53] the number of fa is different the size of the pick is different so there are so
[01:55:55] of the pick is different so there are so many variable forget about the
[01:55:57] many variable forget about the technology the variable in the package
[01:55:59] technology the variable in the package itself so it's very difficult for oset
[01:56:02] itself so it's very difficult for oset or the CMS to invest because the
[01:56:05] or the CMS to invest because the customer itself is keep changing the
[01:56:07] customer itself is keep changing the device size the form factor the
[01:56:10] device size the form factor the thickness the FAU so I think it's very
[01:56:12] thickness the FAU so I think it's very difficult to invest U big time uh to
[01:56:15] difficult to invest U big time uh to ramp up the capacity till customer says
[01:56:18] ramp up the capacity till customer says you know what I'm going to have a fully
[01:56:20] you know what I'm going to have a fully consigned uh cap capacity for myself and
[01:56:24] consigned uh cap capacity for myself and they are responsible for it that's one
[01:56:26] they are responsible for it that's one of the reason people like AC or mcore
[01:56:29] of the reason people like AC or mcore not able to step up because I think it
[01:56:32] not able to step up because I think it is another in my opinion three to five
[01:56:34] is another in my opinion three to five years before the industri you know have
[01:56:37] years before the industri you know have the optical package size stable and then
[01:56:40] the optical package size stable and then people like f cont or whoever can step
[01:56:43] people like f cont or whoever can step up and then bring in the equipment and
[01:56:44] up and then bring in the equipment and we can have the true volume in millions
[01:56:47] we can have the true volume in millions or billions of unit you know per year
[01:56:50] or billions of unit you know per year so
[01:56:52] so question I want to go back to broadcom
[01:56:54] question I want to go back to broadcom now we have a lot of questions in the
[01:56:56] now we have a lot of questions in the room about chiplets everybody's talking
[01:56:59] room about chiplets everybody's talking about chiplets these days and they
[01:57:00] about chiplets these days and they consider that the role of photonis will
[01:57:02] consider that the role of photonis will be the interconnection between chiplets
[01:57:05] be the interconnection between chiplets H from the broad con side do you see the
[01:57:07] H from the broad con side do you see the need to go to certain level of
[01:57:10] need to go to certain level of standardization to drive the optical
[01:57:12] standardization to drive the optical chiplet in
[01:57:14] chiplet in interconnect so so we so we've started
[01:57:19] interconnect so so we so we've started by looking number one at applications
[01:57:22] by looking number one at applications that are inherently Optical to begin
[01:57:24] that are inherently Optical to begin with so a lot of what's emerging let's
[01:57:27] with so a lot of what's emerging let's say for sort of the Next Generation
[01:57:29] say for sort of the Next Generation shiplet interconnect for Optics um are
[01:57:32] shiplet interconnect for Optics um are looking at some of the applications that
[01:57:34] looking at some of the applications that we have not been targeting so in this
[01:57:36] we have not been targeting so in this case the interconnect say between your
[01:57:38] case the interconnect say between your Asic you know like take CPO for example
[01:57:40] Asic you know like take CPO for example which is inherently a chiplet
[01:57:42] which is inherently a chiplet architecture but uh you're using a
[01:57:44] architecture but uh you're using a traditional connection of say an LR CES
[01:57:47] traditional connection of say an LR CES or something of that that nature for the
[01:57:49] or something of that that nature for the the connection I think for networking
[01:57:51] the connection I think for networking applications that is is still the um
[01:57:55] applications that is is still the um methodology that that we see kind of
[01:57:57] methodology that that we see kind of persisting over the next few
[01:58:00] persisting over the next few Generations um I would not be the right
[01:58:03] Generations um I would not be the right one to comment on what we may see after
[01:58:06] one to comment on what we may see after that for some of the more Advanced
[01:58:08] that for some of the more Advanced Computer Applications once you start do
[01:58:10] Computer Applications once you start do it trying to do chipto chip Optical you
[01:58:12] it trying to do chipto chip Optical you know connectivity on a die Level or on a
[01:58:14] know connectivity on a die Level or on a package level so there are there are
[01:58:16] package level so there are there are many many more questions in the chat
[01:58:18] many many more questions in the chat allow me not to continue with them many
[01:58:21] allow me not to continue with them many questions about Acquisitions many
[01:58:23] questions about Acquisitions many questions that I'm not going to address
[01:58:25] questions that I'm not going to address on this but I'm going to encourage all
[01:58:27] on this but I'm going to encourage all the people here to answer the question
[01:58:29] the people here to answer the question themselves in the chat so everybody can
[01:58:31] themselves in the chat so everybody can get their answer but now there is time
[01:58:33] get their answer but now there is time to get the floor to a very dear friend
[01:58:36] to get the floor to a very dear friend of mine a person who has m a lot in my
[01:58:38] of mine a person who has m a lot in my career Melissa caros thank you very much
[01:58:41] career Melissa caros thank you very much for this opportunity of sharing this
[01:58:43] for this opportunity of sharing this webcast of the advanced photonic
[01:58:44] webcast of the advanced photonic Coalition I would like to give you the
[01:58:46] Coalition I would like to give you the floor to tell us what's next with APC
[01:58:52] I can tell you it's collaboration
[01:58:54] I can tell you it's collaboration without us all collaborating as an
[01:58:57] without us all collaborating as an industry to move forward the challenges
[01:58:59] industry to move forward the challenges that we have uh to overcome any design
[01:59:03] that we have uh to overcome any design or
[01:59:04] or manufacturing uh
[01:59:06] manufacturing uh situation uh it's all of us working
[01:59:09] situation uh it's all of us working together towards a common goal and uh
[01:59:12] together towards a common goal and uh that's really it it's collaboration to
[01:59:15] that's really it it's collaboration to overcome our current challenges and meet
[01:59:18] overcome our current challenges and meet uh the needs of this industry thank you
[01:59:20] uh the needs of this industry thank you and of course all of us are going to San
[01:59:22] and of course all of us are going to San Diego for ofc Jeff Jeff freaki what's
[01:59:26] Diego for ofc Jeff Jeff freaki what's going to happen with APC during
[01:59:30] going to happen with APC during ofc we will have a panel uh in in the
[01:59:35] ofc we will have a panel uh in in the trade show floor and the panelists will
[01:59:39] trade show floor and the panelists will be further uh exploring what's happening
[01:59:41] be further uh exploring what's happening in this area so we haven't completely
[01:59:44] in this area so we haven't completely finalized the uh the pro the content of
[01:59:48] finalized the uh the pro the content of the panel but we'll be there and I will
[01:59:51] the panel but we'll be there and I will be there as well of course ofc is the
[01:59:53] be there as well of course ofc is the place to be but it's still two months
[01:59:55] place to be but it's still two months ahead if you have any questions for APC
[01:59:58] ahead if you have any questions for APC contact Melissa Carlos and Jeffrey maky
[02:00:00] contact Melissa Carlos and Jeffrey maky you have any questions about Optica
[02:00:01] you have any questions about Optica contact me but most important if you
[02:00:03] contact me but most important if you have any questions about the supply
[02:00:05] have any questions about the supply chain of photonics for AI contact Manish
[02:00:09] chain of photonics for AI contact Manish mea from broadcom P SRA from quantify
[02:00:12] mea from broadcom P SRA from quantify vas SCA from Global foundaries Das
[02:00:15] vas SCA from Global foundaries Das shanandra Tronics and D baram from F
[02:00:19] shanandra Tronics and D baram from F cont Tech it has been an honor and a
[02:00:22] cont Tech it has been an honor and a pleasure being with you today morning
[02:00:24] pleasure being with you today morning afternoon and evening in the world it
[02:00:25] afternoon and evening in the world it was a global event thank you until the
[02:00:27] was a global event thank you until the next time take care of each other if you
[02:00:29] next time take care of each other if you have to wear a mask recover from your
[02:00:31] have to wear a mask recover from your cult and see you at ofc
[02:00:35] cult and see you at ofc byebye thank you for
[02:00:41] [Music]
[02:00:44] [Music] watching

https://www.youtube.com/watch?v=ke4Dqcjl3lo

Summary

TL;DR — The Advanced Photonics Coalition (APC) is fostering collaboration and standardization within the integrated photonics industry to meet the demands of the AI revolution. Key players across the entire value chain, from foundries to equipment manufacturers and end-users, are working together to address challenges in scaling production, reducing costs, and ensuring reliability. The event highlighted the critical role of silicon photonics in enabling high-speed, energy-efficient communication for AI and HPC applications.

Key points

• The APC aims to create a neutral ground for industry collaboration, standardize technologies, and accelerate the scaling of photonic integrated circuits (PICs) to high volumes.

• Foundries like GlobalFoundries offer flexible silicon photonics platforms with pre-characterized devices, enabling various architectures for high-speed interconnects.

• OSATs and CMs face challenges in adapting to the unique demands of photonic assembly, including tighter tolerances, diverse materials, and the need for specialized equipment and processes.

• Equipment manufacturers are developing advanced solutions for wafer-level testing, high-precision die bonding, and automated fiber handling to support high-volume production.

• Co-packaged optics (CPO) and remote laser modules are emerging as key technologies for improving power efficiency and density in AI data centers, though standardization remains a challenge.

• The ecosystem requires close collaboration between EDA tool vendors, foundries, OSATs, CMs, IP providers, and equipment suppliers to overcome hurdles and unlock the full potential of silicon photonics.

Takeaway — Standardization and cross-ecosystem collaboration are essential for the integrated photonics industry to successfully enable the AI revolution and achieve high-volume, cost-effective production.